]>
Commit | Line | Data |
---|---|---|
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 | */ | |
0d75590d | 27 | #include "qemu/osdep.h" |
da34e65c | 28 | #include "qapi/error.h" |
9c17d615 | 29 | #include "sysemu/sysemu.h" |
e35704ba | 30 | #include "sysemu/numa.h" |
83c9f4ca | 31 | #include "hw/hw.h" |
03dd024f | 32 | #include "qemu/log.h" |
71461b0f | 33 | #include "hw/fw-path-provider.h" |
9fdf0c29 | 34 | #include "elf.h" |
1422e32d | 35 | #include "net/net.h" |
ad440b4a | 36 | #include "sysemu/device_tree.h" |
fa1d36df | 37 | #include "sysemu/block-backend.h" |
9c17d615 | 38 | #include "sysemu/cpus.h" |
b3946626 | 39 | #include "sysemu/hw_accel.h" |
e97c3636 | 40 | #include "kvm_ppc.h" |
ff14e817 | 41 | #include "migration/migration.h" |
f2a8f0a6 | 42 | #include "migration/register.h" |
4be21d56 | 43 | #include "mmu-hash64.h" |
b4db5413 | 44 | #include "mmu-book3s-v3.h" |
3794d548 | 45 | #include "qom/cpu.h" |
9fdf0c29 DG |
46 | |
47 | #include "hw/boards.h" | |
0d09e41a | 48 | #include "hw/ppc/ppc.h" |
9fdf0c29 DG |
49 | #include "hw/loader.h" |
50 | ||
7804c353 | 51 | #include "hw/ppc/fdt.h" |
0d09e41a PB |
52 | #include "hw/ppc/spapr.h" |
53 | #include "hw/ppc/spapr_vio.h" | |
54 | #include "hw/pci-host/spapr.h" | |
55 | #include "hw/ppc/xics.h" | |
a2cb15b0 | 56 | #include "hw/pci/msi.h" |
9fdf0c29 | 57 | |
83c9f4ca | 58 | #include "hw/pci/pci.h" |
71461b0f AK |
59 | #include "hw/scsi/scsi.h" |
60 | #include "hw/virtio/virtio-scsi.h" | |
c4e13492 | 61 | #include "hw/virtio/vhost-scsi-common.h" |
f61b4bed | 62 | |
022c62cb | 63 | #include "exec/address-spaces.h" |
35139a59 | 64 | #include "hw/usb.h" |
1de7afc9 | 65 | #include "qemu/config-file.h" |
135a129a | 66 | #include "qemu/error-report.h" |
2a6593cb | 67 | #include "trace.h" |
34316482 | 68 | #include "hw/nmi.h" |
6449da45 | 69 | #include "hw/intc/intc.h" |
890c2b77 | 70 | |
68a27b20 | 71 | #include "hw/compat.h" |
f348b6d1 | 72 | #include "qemu/cutils.h" |
94a94e4c | 73 | #include "hw/ppc/spapr_cpu_core.h" |
2474bfd4 | 74 | #include "qmp-commands.h" |
68a27b20 | 75 | |
9fdf0c29 DG |
76 | #include <libfdt.h> |
77 | ||
4d8d5467 BH |
78 | /* SLOF memory layout: |
79 | * | |
80 | * SLOF raw image loaded at 0, copies its romfs right below the flat | |
81 | * device-tree, then position SLOF itself 31M below that | |
82 | * | |
83 | * So we set FW_OVERHEAD to 40MB which should account for all of that | |
84 | * and more | |
85 | * | |
86 | * We load our kernel at 4M, leaving space for SLOF initial image | |
87 | */ | |
38b02bd8 | 88 | #define FDT_MAX_SIZE 0x100000 |
39ac8455 | 89 | #define RTAS_MAX_SIZE 0x10000 |
b7d1f77a | 90 | #define RTAS_MAX_ADDR 0x80000000 /* RTAS must stay below that */ |
a9f8ad8f DG |
91 | #define FW_MAX_SIZE 0x400000 |
92 | #define FW_FILE_NAME "slof.bin" | |
4d8d5467 BH |
93 | #define FW_OVERHEAD 0x2800000 |
94 | #define KERNEL_LOAD_ADDR FW_MAX_SIZE | |
a9f8ad8f | 95 | |
4d8d5467 | 96 | #define MIN_RMA_SLOF 128UL |
9fdf0c29 | 97 | |
0c103f8e DG |
98 | #define PHANDLE_XICP 0x00001111 |
99 | ||
7f763a5d DG |
100 | #define HTAB_SIZE(spapr) (1ULL << ((spapr)->htab_shift)) |
101 | ||
71cd4dac CLG |
102 | static ICSState *spapr_ics_create(sPAPRMachineState *spapr, |
103 | const char *type_ics, | |
104 | int nr_irqs, Error **errp) | |
c04d6cfa | 105 | { |
175d2aa0 | 106 | Error *local_err = NULL; |
71cd4dac | 107 | Object *obj; |
4e4169f7 | 108 | |
71cd4dac | 109 | obj = object_new(type_ics); |
175d2aa0 | 110 | object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); |
71cd4dac | 111 | object_property_add_const_link(obj, "xics", OBJECT(spapr), &error_abort); |
175d2aa0 GK |
112 | object_property_set_int(obj, nr_irqs, "nr-irqs", &local_err); |
113 | if (local_err) { | |
114 | goto error; | |
115 | } | |
71cd4dac | 116 | object_property_set_bool(obj, true, "realized", &local_err); |
175d2aa0 GK |
117 | if (local_err) { |
118 | goto error; | |
4e4169f7 | 119 | } |
4e4169f7 | 120 | |
71cd4dac | 121 | return ICS_SIMPLE(obj); |
175d2aa0 GK |
122 | |
123 | error: | |
124 | error_propagate(errp, local_err); | |
125 | return NULL; | |
c04d6cfa AL |
126 | } |
127 | ||
71cd4dac | 128 | static void xics_system_init(MachineState *machine, int nr_irqs, Error **errp) |
c04d6cfa | 129 | { |
71cd4dac | 130 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
c04d6cfa | 131 | |
11ad93f6 | 132 | if (kvm_enabled()) { |
2192a930 | 133 | if (machine_kernel_irqchip_allowed(machine) && |
71cd4dac CLG |
134 | !xics_kvm_init(spapr, errp)) { |
135 | spapr->icp_type = TYPE_KVM_ICP; | |
3d85885a | 136 | spapr->ics = spapr_ics_create(spapr, TYPE_ICS_KVM, nr_irqs, errp); |
11ad93f6 | 137 | } |
71cd4dac | 138 | if (machine_kernel_irqchip_required(machine) && !spapr->ics) { |
3d85885a GK |
139 | error_prepend(errp, "kernel_irqchip requested but unavailable: "); |
140 | return; | |
11ad93f6 DG |
141 | } |
142 | } | |
143 | ||
71cd4dac | 144 | if (!spapr->ics) { |
f63ebfe0 | 145 | xics_spapr_init(spapr); |
71cd4dac CLG |
146 | spapr->icp_type = TYPE_ICP; |
147 | spapr->ics = spapr_ics_create(spapr, TYPE_ICS_SIMPLE, nr_irqs, errp); | |
3d85885a GK |
148 | if (!spapr->ics) { |
149 | return; | |
150 | } | |
c04d6cfa | 151 | } |
c04d6cfa AL |
152 | } |
153 | ||
833d4668 AK |
154 | static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu, |
155 | int smt_threads) | |
156 | { | |
157 | int i, ret = 0; | |
158 | uint32_t servers_prop[smt_threads]; | |
159 | uint32_t gservers_prop[smt_threads * 2]; | |
160 | int index = ppc_get_vcpu_dt_id(cpu); | |
161 | ||
d6e166c0 DG |
162 | if (cpu->compat_pvr) { |
163 | ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->compat_pvr); | |
6d9412ea AK |
164 | if (ret < 0) { |
165 | return ret; | |
166 | } | |
167 | } | |
168 | ||
833d4668 AK |
169 | /* Build interrupt servers and gservers properties */ |
170 | for (i = 0; i < smt_threads; i++) { | |
171 | servers_prop[i] = cpu_to_be32(index + i); | |
172 | /* Hack, direct the group queues back to cpu 0 */ | |
173 | gservers_prop[i*2] = cpu_to_be32(index + i); | |
174 | gservers_prop[i*2 + 1] = 0; | |
175 | } | |
176 | ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s", | |
177 | servers_prop, sizeof(servers_prop)); | |
178 | if (ret < 0) { | |
179 | return ret; | |
180 | } | |
181 | ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s", | |
182 | gservers_prop, sizeof(gservers_prop)); | |
183 | ||
184 | return ret; | |
185 | } | |
186 | ||
99861ecb | 187 | static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, PowerPCCPU *cpu) |
0da6f3fe | 188 | { |
0da6f3fe BR |
189 | int index = ppc_get_vcpu_dt_id(cpu); |
190 | uint32_t associativity[] = {cpu_to_be32(0x5), | |
191 | cpu_to_be32(0x0), | |
192 | cpu_to_be32(0x0), | |
193 | cpu_to_be32(0x0), | |
15f8b142 | 194 | cpu_to_be32(cpu->node_id), |
0da6f3fe BR |
195 | cpu_to_be32(index)}; |
196 | ||
197 | /* Advertise NUMA via ibm,associativity */ | |
99861ecb | 198 | return fdt_setprop(fdt, offset, "ibm,associativity", associativity, |
0da6f3fe | 199 | sizeof(associativity)); |
0da6f3fe BR |
200 | } |
201 | ||
86d5771a | 202 | /* Populate the "ibm,pa-features" property */ |
e957f6a9 SB |
203 | static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset, |
204 | bool legacy_guest) | |
86d5771a SB |
205 | { |
206 | uint8_t pa_features_206[] = { 6, 0, | |
207 | 0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 }; | |
208 | uint8_t pa_features_207[] = { 24, 0, | |
209 | 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0, | |
210 | 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, | |
211 | 0x00, 0x00, 0x00, 0x00, 0x80, 0x00, | |
212 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00 }; | |
9fb4541f SB |
213 | uint8_t pa_features_300[] = { 66, 0, |
214 | /* 0: MMU|FPU|SLB|RUN|DABR|NX, 1: fri[nzpm]|DABRX|SPRG3|SLB0|PP110 */ | |
215 | /* 2: VPM|DS205|PPR|DS202|DS206, 3: LSD|URG, SSO, 5: LE|CFAR|EB|LSQ */ | |
216 | 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0, /* 0 - 5 */ | |
217 | /* 6: DS207 */ | |
218 | 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, /* 6 - 11 */ | |
219 | /* 16: Vector */ | |
86d5771a | 220 | 0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 12 - 17 */ |
9fb4541f | 221 | /* 18: Vec. Scalar, 20: Vec. XOR, 22: HTM */ |
9bf502fe | 222 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 18 - 23 */ |
9fb4541f SB |
223 | /* 24: Ext. Dec, 26: 64 bit ftrs, 28: PM ftrs */ |
224 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 24 - 29 */ | |
225 | /* 30: MMR, 32: LE atomic, 34: EBB + ext EBB */ | |
226 | 0x80, 0x00, 0x80, 0x00, 0xC0, 0x00, /* 30 - 35 */ | |
227 | /* 36: SPR SO, 38: Copy/Paste, 40: Radix MMU */ | |
228 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 36 - 41 */ | |
229 | /* 42: PM, 44: PC RA, 46: SC vec'd */ | |
230 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 42 - 47 */ | |
231 | /* 48: SIMD, 50: QP BFP, 52: String */ | |
232 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 48 - 53 */ | |
233 | /* 54: DecFP, 56: DecI, 58: SHA */ | |
234 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 54 - 59 */ | |
235 | /* 60: NM atomic, 62: RNG */ | |
236 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 60 - 65 */ | |
237 | }; | |
86d5771a SB |
238 | uint8_t *pa_features; |
239 | size_t pa_size; | |
240 | ||
241 | switch (POWERPC_MMU_VER(env->mmu_model)) { | |
242 | case POWERPC_MMU_VER_2_06: | |
243 | pa_features = pa_features_206; | |
244 | pa_size = sizeof(pa_features_206); | |
245 | break; | |
246 | case POWERPC_MMU_VER_2_07: | |
247 | pa_features = pa_features_207; | |
248 | pa_size = sizeof(pa_features_207); | |
249 | break; | |
250 | case POWERPC_MMU_VER_3_00: | |
251 | pa_features = pa_features_300; | |
252 | pa_size = sizeof(pa_features_300); | |
253 | break; | |
254 | default: | |
255 | return; | |
256 | } | |
257 | ||
258 | if (env->ci_large_pages) { | |
259 | /* | |
260 | * Note: we keep CI large pages off by default because a 64K capable | |
261 | * guest provisioned with large pages might otherwise try to map a qemu | |
262 | * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages | |
263 | * even if that qemu runs on a 4k host. | |
264 | * We dd this bit back here if we are confident this is not an issue | |
265 | */ | |
266 | pa_features[3] |= 0x20; | |
267 | } | |
268 | if (kvmppc_has_cap_htm() && pa_size > 24) { | |
269 | pa_features[24] |= 0x80; /* Transactional memory support */ | |
270 | } | |
e957f6a9 SB |
271 | if (legacy_guest && pa_size > 40) { |
272 | /* Workaround for broken kernels that attempt (guest) radix | |
273 | * mode when they can't handle it, if they see the radix bit set | |
274 | * in pa-features. So hide it from them. */ | |
275 | pa_features[40 + 2] &= ~0x80; /* Radix MMU */ | |
276 | } | |
86d5771a SB |
277 | |
278 | _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size))); | |
279 | } | |
280 | ||
28e02042 | 281 | static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr) |
6e806cc3 | 282 | { |
82677ed2 AK |
283 | int ret = 0, offset, cpus_offset; |
284 | CPUState *cs; | |
6e806cc3 BR |
285 | char cpu_model[32]; |
286 | int smt = kvmppc_smt_threads(); | |
7f763a5d | 287 | uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; |
6e806cc3 | 288 | |
82677ed2 AK |
289 | CPU_FOREACH(cs) { |
290 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
e957f6a9 | 291 | CPUPPCState *env = &cpu->env; |
82677ed2 AK |
292 | DeviceClass *dc = DEVICE_GET_CLASS(cs); |
293 | int index = ppc_get_vcpu_dt_id(cpu); | |
12dbeb16 | 294 | int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); |
6e806cc3 | 295 | |
0f20ba62 | 296 | if ((index % smt) != 0) { |
6e806cc3 BR |
297 | continue; |
298 | } | |
299 | ||
82677ed2 | 300 | snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index); |
6e806cc3 | 301 | |
82677ed2 AK |
302 | cpus_offset = fdt_path_offset(fdt, "/cpus"); |
303 | if (cpus_offset < 0) { | |
304 | cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"), | |
305 | "cpus"); | |
306 | if (cpus_offset < 0) { | |
307 | return cpus_offset; | |
308 | } | |
309 | } | |
310 | offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model); | |
6e806cc3 | 311 | if (offset < 0) { |
82677ed2 AK |
312 | offset = fdt_add_subnode(fdt, cpus_offset, cpu_model); |
313 | if (offset < 0) { | |
314 | return offset; | |
315 | } | |
6e806cc3 BR |
316 | } |
317 | ||
7f763a5d DG |
318 | ret = fdt_setprop(fdt, offset, "ibm,pft-size", |
319 | pft_size_prop, sizeof(pft_size_prop)); | |
6e806cc3 BR |
320 | if (ret < 0) { |
321 | return ret; | |
322 | } | |
833d4668 | 323 | |
99861ecb IM |
324 | if (nb_numa_nodes > 1) { |
325 | ret = spapr_fixup_cpu_numa_dt(fdt, offset, cpu); | |
326 | if (ret < 0) { | |
327 | return ret; | |
328 | } | |
0da6f3fe BR |
329 | } |
330 | ||
12dbeb16 | 331 | ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt); |
833d4668 AK |
332 | if (ret < 0) { |
333 | return ret; | |
334 | } | |
e957f6a9 SB |
335 | |
336 | spapr_populate_pa_features(env, fdt, offset, | |
337 | spapr->cas_legacy_guest_workaround); | |
6e806cc3 BR |
338 | } |
339 | return ret; | |
340 | } | |
341 | ||
b082d65a AK |
342 | static hwaddr spapr_node0_size(void) |
343 | { | |
fb164994 DG |
344 | MachineState *machine = MACHINE(qdev_get_machine()); |
345 | ||
b082d65a AK |
346 | if (nb_numa_nodes) { |
347 | int i; | |
348 | for (i = 0; i < nb_numa_nodes; ++i) { | |
349 | if (numa_info[i].node_mem) { | |
fb164994 DG |
350 | return MIN(pow2floor(numa_info[i].node_mem), |
351 | machine->ram_size); | |
b082d65a AK |
352 | } |
353 | } | |
354 | } | |
fb164994 | 355 | return machine->ram_size; |
b082d65a AK |
356 | } |
357 | ||
a1d59c0f AK |
358 | static void add_str(GString *s, const gchar *s1) |
359 | { | |
360 | g_string_append_len(s, s1, strlen(s1) + 1); | |
361 | } | |
7f763a5d | 362 | |
03d196b7 | 363 | static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start, |
26a8c353 AK |
364 | hwaddr size) |
365 | { | |
366 | uint32_t associativity[] = { | |
367 | cpu_to_be32(0x4), /* length */ | |
368 | cpu_to_be32(0x0), cpu_to_be32(0x0), | |
c3b4f589 | 369 | cpu_to_be32(0x0), cpu_to_be32(nodeid) |
26a8c353 AK |
370 | }; |
371 | char mem_name[32]; | |
372 | uint64_t mem_reg_property[2]; | |
373 | int off; | |
374 | ||
375 | mem_reg_property[0] = cpu_to_be64(start); | |
376 | mem_reg_property[1] = cpu_to_be64(size); | |
377 | ||
378 | sprintf(mem_name, "memory@" TARGET_FMT_lx, start); | |
379 | off = fdt_add_subnode(fdt, 0, mem_name); | |
380 | _FDT(off); | |
381 | _FDT((fdt_setprop_string(fdt, off, "device_type", "memory"))); | |
382 | _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property, | |
383 | sizeof(mem_reg_property)))); | |
384 | _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity, | |
385 | sizeof(associativity)))); | |
03d196b7 | 386 | return off; |
26a8c353 AK |
387 | } |
388 | ||
28e02042 | 389 | static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt) |
7f763a5d | 390 | { |
fb164994 | 391 | MachineState *machine = MACHINE(spapr); |
7db8a127 AK |
392 | hwaddr mem_start, node_size; |
393 | int i, nb_nodes = nb_numa_nodes; | |
394 | NodeInfo *nodes = numa_info; | |
395 | NodeInfo ramnode; | |
396 | ||
397 | /* No NUMA nodes, assume there is just one node with whole RAM */ | |
398 | if (!nb_numa_nodes) { | |
399 | nb_nodes = 1; | |
fb164994 | 400 | ramnode.node_mem = machine->ram_size; |
7db8a127 | 401 | nodes = &ramnode; |
5fe269b1 | 402 | } |
7f763a5d | 403 | |
7db8a127 AK |
404 | for (i = 0, mem_start = 0; i < nb_nodes; ++i) { |
405 | if (!nodes[i].node_mem) { | |
406 | continue; | |
407 | } | |
fb164994 | 408 | if (mem_start >= machine->ram_size) { |
5fe269b1 PM |
409 | node_size = 0; |
410 | } else { | |
7db8a127 | 411 | node_size = nodes[i].node_mem; |
fb164994 DG |
412 | if (node_size > machine->ram_size - mem_start) { |
413 | node_size = machine->ram_size - mem_start; | |
5fe269b1 PM |
414 | } |
415 | } | |
7db8a127 AK |
416 | if (!mem_start) { |
417 | /* ppc_spapr_init() checks for rma_size <= node0_size already */ | |
e8f986fc | 418 | spapr_populate_memory_node(fdt, i, 0, spapr->rma_size); |
7db8a127 AK |
419 | mem_start += spapr->rma_size; |
420 | node_size -= spapr->rma_size; | |
421 | } | |
6010818c AK |
422 | for ( ; node_size; ) { |
423 | hwaddr sizetmp = pow2floor(node_size); | |
424 | ||
425 | /* mem_start != 0 here */ | |
426 | if (ctzl(mem_start) < ctzl(sizetmp)) { | |
427 | sizetmp = 1ULL << ctzl(mem_start); | |
428 | } | |
429 | ||
430 | spapr_populate_memory_node(fdt, i, mem_start, sizetmp); | |
431 | node_size -= sizetmp; | |
432 | mem_start += sizetmp; | |
433 | } | |
7f763a5d DG |
434 | } |
435 | ||
436 | return 0; | |
437 | } | |
438 | ||
0da6f3fe BR |
439 | static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset, |
440 | sPAPRMachineState *spapr) | |
441 | { | |
442 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
443 | CPUPPCState *env = &cpu->env; | |
444 | PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs); | |
445 | int index = ppc_get_vcpu_dt_id(cpu); | |
446 | uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40), | |
447 | 0xffffffff, 0xffffffff}; | |
afd10a0f BR |
448 | uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() |
449 | : SPAPR_TIMEBASE_FREQ; | |
0da6f3fe BR |
450 | uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000; |
451 | uint32_t page_sizes_prop[64]; | |
452 | size_t page_sizes_prop_size; | |
22419c2a | 453 | uint32_t vcpus_per_socket = smp_threads * smp_cores; |
0da6f3fe | 454 | uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; |
12dbeb16 | 455 | int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); |
af81cf32 | 456 | sPAPRDRConnector *drc; |
af81cf32 | 457 | int drc_index; |
c64abd1f SB |
458 | uint32_t radix_AP_encodings[PPC_PAGE_SIZES_MAX_SZ]; |
459 | int i; | |
af81cf32 | 460 | |
fbf55397 | 461 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index); |
af81cf32 | 462 | if (drc) { |
0b55aa91 | 463 | drc_index = spapr_drc_index(drc); |
af81cf32 BR |
464 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index))); |
465 | } | |
0da6f3fe BR |
466 | |
467 | _FDT((fdt_setprop_cell(fdt, offset, "reg", index))); | |
468 | _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu"))); | |
469 | ||
470 | _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR]))); | |
471 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size", | |
472 | env->dcache_line_size))); | |
473 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size", | |
474 | env->dcache_line_size))); | |
475 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size", | |
476 | env->icache_line_size))); | |
477 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size", | |
478 | env->icache_line_size))); | |
479 | ||
480 | if (pcc->l1_dcache_size) { | |
481 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size", | |
482 | pcc->l1_dcache_size))); | |
483 | } else { | |
ce9863b7 | 484 | error_report("Warning: Unknown L1 dcache size for cpu"); |
0da6f3fe BR |
485 | } |
486 | if (pcc->l1_icache_size) { | |
487 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size", | |
488 | pcc->l1_icache_size))); | |
489 | } else { | |
ce9863b7 | 490 | error_report("Warning: Unknown L1 icache size for cpu"); |
0da6f3fe BR |
491 | } |
492 | ||
493 | _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq))); | |
494 | _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq))); | |
fd5da5c4 | 495 | _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr))); |
0da6f3fe BR |
496 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr))); |
497 | _FDT((fdt_setprop_string(fdt, offset, "status", "okay"))); | |
498 | _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0))); | |
499 | ||
500 | if (env->spr_cb[SPR_PURR].oea_read) { | |
501 | _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0))); | |
502 | } | |
503 | ||
504 | if (env->mmu_model & POWERPC_MMU_1TSEG) { | |
505 | _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes", | |
506 | segs, sizeof(segs)))); | |
507 | } | |
508 | ||
509 | /* Advertise VMX/VSX (vector extensions) if available | |
510 | * 0 / no property == no vector extensions | |
511 | * 1 == VMX / Altivec available | |
512 | * 2 == VSX available */ | |
513 | if (env->insns_flags & PPC_ALTIVEC) { | |
514 | uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1; | |
515 | ||
516 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx))); | |
517 | } | |
518 | ||
519 | /* Advertise DFP (Decimal Floating Point) if available | |
520 | * 0 / no property == no DFP | |
521 | * 1 == DFP available */ | |
522 | if (env->insns_flags2 & PPC2_DFP) { | |
523 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1))); | |
524 | } | |
525 | ||
3654fa95 | 526 | page_sizes_prop_size = ppc_create_page_sizes_prop(env, page_sizes_prop, |
0da6f3fe BR |
527 | sizeof(page_sizes_prop)); |
528 | if (page_sizes_prop_size) { | |
529 | _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes", | |
530 | page_sizes_prop, page_sizes_prop_size))); | |
531 | } | |
532 | ||
e957f6a9 | 533 | spapr_populate_pa_features(env, fdt, offset, false); |
90da0d5a | 534 | |
0da6f3fe | 535 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", |
22419c2a | 536 | cs->cpu_index / vcpus_per_socket))); |
0da6f3fe BR |
537 | |
538 | _FDT((fdt_setprop(fdt, offset, "ibm,pft-size", | |
539 | pft_size_prop, sizeof(pft_size_prop)))); | |
540 | ||
99861ecb IM |
541 | if (nb_numa_nodes > 1) { |
542 | _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cpu)); | |
543 | } | |
0da6f3fe | 544 | |
12dbeb16 | 545 | _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt)); |
c64abd1f SB |
546 | |
547 | if (pcc->radix_page_info) { | |
548 | for (i = 0; i < pcc->radix_page_info->count; i++) { | |
549 | radix_AP_encodings[i] = | |
550 | cpu_to_be32(pcc->radix_page_info->entries[i]); | |
551 | } | |
552 | _FDT((fdt_setprop(fdt, offset, "ibm,processor-radix-AP-encodings", | |
553 | radix_AP_encodings, | |
554 | pcc->radix_page_info->count * | |
555 | sizeof(radix_AP_encodings[0])))); | |
556 | } | |
0da6f3fe BR |
557 | } |
558 | ||
559 | static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr) | |
560 | { | |
561 | CPUState *cs; | |
562 | int cpus_offset; | |
563 | char *nodename; | |
564 | int smt = kvmppc_smt_threads(); | |
565 | ||
566 | cpus_offset = fdt_add_subnode(fdt, 0, "cpus"); | |
567 | _FDT(cpus_offset); | |
568 | _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1))); | |
569 | _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0))); | |
570 | ||
571 | /* | |
572 | * We walk the CPUs in reverse order to ensure that CPU DT nodes | |
573 | * created by fdt_add_subnode() end up in the right order in FDT | |
574 | * for the guest kernel the enumerate the CPUs correctly. | |
575 | */ | |
576 | CPU_FOREACH_REVERSE(cs) { | |
577 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
578 | int index = ppc_get_vcpu_dt_id(cpu); | |
579 | DeviceClass *dc = DEVICE_GET_CLASS(cs); | |
580 | int offset; | |
581 | ||
582 | if ((index % smt) != 0) { | |
583 | continue; | |
584 | } | |
585 | ||
586 | nodename = g_strdup_printf("%s@%x", dc->fw_name, index); | |
587 | offset = fdt_add_subnode(fdt, cpus_offset, nodename); | |
588 | g_free(nodename); | |
589 | _FDT(offset); | |
590 | spapr_populate_cpu_dt(cs, fdt, offset, spapr); | |
591 | } | |
592 | ||
593 | } | |
594 | ||
03d196b7 BR |
595 | /* |
596 | * Adds ibm,dynamic-reconfiguration-memory node. | |
597 | * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation | |
598 | * of this device tree node. | |
599 | */ | |
600 | static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt) | |
601 | { | |
602 | MachineState *machine = MACHINE(spapr); | |
603 | int ret, i, offset; | |
604 | uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; | |
605 | uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)}; | |
d0e5a8f2 BR |
606 | uint32_t hotplug_lmb_start = spapr->hotplug_memory.base / lmb_size; |
607 | uint32_t nr_lmbs = (spapr->hotplug_memory.base + | |
608 | memory_region_size(&spapr->hotplug_memory.mr)) / | |
609 | lmb_size; | |
03d196b7 | 610 | uint32_t *int_buf, *cur_index, buf_len; |
6663864e | 611 | int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1; |
03d196b7 | 612 | |
16c25aef | 613 | /* |
d0e5a8f2 | 614 | * Don't create the node if there is no hotpluggable memory |
16c25aef | 615 | */ |
d0e5a8f2 | 616 | if (machine->ram_size == machine->maxram_size) { |
16c25aef BR |
617 | return 0; |
618 | } | |
619 | ||
ef001f06 TH |
620 | /* |
621 | * Allocate enough buffer size to fit in ibm,dynamic-memory | |
622 | * or ibm,associativity-lookup-arrays | |
623 | */ | |
624 | buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2) | |
625 | * sizeof(uint32_t); | |
03d196b7 BR |
626 | cur_index = int_buf = g_malloc0(buf_len); |
627 | ||
628 | offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory"); | |
629 | ||
630 | ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size, | |
631 | sizeof(prop_lmb_size)); | |
632 | if (ret < 0) { | |
633 | goto out; | |
634 | } | |
635 | ||
636 | ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff); | |
637 | if (ret < 0) { | |
638 | goto out; | |
639 | } | |
640 | ||
641 | ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0); | |
642 | if (ret < 0) { | |
643 | goto out; | |
644 | } | |
645 | ||
646 | /* ibm,dynamic-memory */ | |
647 | int_buf[0] = cpu_to_be32(nr_lmbs); | |
648 | cur_index++; | |
649 | for (i = 0; i < nr_lmbs; i++) { | |
d0e5a8f2 | 650 | uint64_t addr = i * lmb_size; |
03d196b7 BR |
651 | uint32_t *dynamic_memory = cur_index; |
652 | ||
d0e5a8f2 BR |
653 | if (i >= hotplug_lmb_start) { |
654 | sPAPRDRConnector *drc; | |
d0e5a8f2 | 655 | |
fbf55397 | 656 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, i); |
d0e5a8f2 | 657 | g_assert(drc); |
d0e5a8f2 BR |
658 | |
659 | dynamic_memory[0] = cpu_to_be32(addr >> 32); | |
660 | dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); | |
0b55aa91 | 661 | dynamic_memory[2] = cpu_to_be32(spapr_drc_index(drc)); |
d0e5a8f2 BR |
662 | dynamic_memory[3] = cpu_to_be32(0); /* reserved */ |
663 | dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL)); | |
664 | if (memory_region_present(get_system_memory(), addr)) { | |
665 | dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED); | |
666 | } else { | |
667 | dynamic_memory[5] = cpu_to_be32(0); | |
668 | } | |
03d196b7 | 669 | } else { |
d0e5a8f2 BR |
670 | /* |
671 | * LMB information for RMA, boot time RAM and gap b/n RAM and | |
672 | * hotplug memory region -- all these are marked as reserved | |
673 | * and as having no valid DRC. | |
674 | */ | |
675 | dynamic_memory[0] = cpu_to_be32(addr >> 32); | |
676 | dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); | |
677 | dynamic_memory[2] = cpu_to_be32(0); | |
678 | dynamic_memory[3] = cpu_to_be32(0); /* reserved */ | |
679 | dynamic_memory[4] = cpu_to_be32(-1); | |
680 | dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_RESERVED | | |
681 | SPAPR_LMB_FLAGS_DRC_INVALID); | |
03d196b7 BR |
682 | } |
683 | ||
684 | cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE; | |
685 | } | |
686 | ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len); | |
687 | if (ret < 0) { | |
688 | goto out; | |
689 | } | |
690 | ||
691 | /* ibm,associativity-lookup-arrays */ | |
692 | cur_index = int_buf; | |
6663864e | 693 | int_buf[0] = cpu_to_be32(nr_nodes); |
03d196b7 BR |
694 | int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */ |
695 | cur_index += 2; | |
6663864e | 696 | for (i = 0; i < nr_nodes; i++) { |
03d196b7 BR |
697 | uint32_t associativity[] = { |
698 | cpu_to_be32(0x0), | |
699 | cpu_to_be32(0x0), | |
700 | cpu_to_be32(0x0), | |
701 | cpu_to_be32(i) | |
702 | }; | |
703 | memcpy(cur_index, associativity, sizeof(associativity)); | |
704 | cur_index += 4; | |
705 | } | |
706 | ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf, | |
707 | (cur_index - int_buf) * sizeof(uint32_t)); | |
708 | out: | |
709 | g_free(int_buf); | |
710 | return ret; | |
711 | } | |
712 | ||
6787d27b MR |
713 | static int spapr_dt_cas_updates(sPAPRMachineState *spapr, void *fdt, |
714 | sPAPROptionVector *ov5_updates) | |
715 | { | |
716 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr); | |
417ece33 | 717 | int ret = 0, offset; |
6787d27b MR |
718 | |
719 | /* Generate ibm,dynamic-reconfiguration-memory node if required */ | |
720 | if (spapr_ovec_test(ov5_updates, OV5_DRCONF_MEMORY)) { | |
721 | g_assert(smc->dr_lmb_enabled); | |
722 | ret = spapr_populate_drconf_memory(spapr, fdt); | |
417ece33 MR |
723 | if (ret) { |
724 | goto out; | |
725 | } | |
6787d27b MR |
726 | } |
727 | ||
417ece33 MR |
728 | offset = fdt_path_offset(fdt, "/chosen"); |
729 | if (offset < 0) { | |
730 | offset = fdt_add_subnode(fdt, 0, "chosen"); | |
731 | if (offset < 0) { | |
732 | return offset; | |
733 | } | |
734 | } | |
735 | ret = spapr_ovec_populate_dt(fdt, offset, spapr->ov5_cas, | |
736 | "ibm,architecture-vec-5"); | |
737 | ||
738 | out: | |
6787d27b MR |
739 | return ret; |
740 | } | |
741 | ||
03d196b7 BR |
742 | int spapr_h_cas_compose_response(sPAPRMachineState *spapr, |
743 | target_ulong addr, target_ulong size, | |
6787d27b | 744 | sPAPROptionVector *ov5_updates) |
03d196b7 BR |
745 | { |
746 | void *fdt, *fdt_skel; | |
747 | sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 }; | |
03d196b7 BR |
748 | |
749 | size -= sizeof(hdr); | |
750 | ||
751 | /* Create sceleton */ | |
752 | fdt_skel = g_malloc0(size); | |
753 | _FDT((fdt_create(fdt_skel, size))); | |
754 | _FDT((fdt_begin_node(fdt_skel, ""))); | |
755 | _FDT((fdt_end_node(fdt_skel))); | |
756 | _FDT((fdt_finish(fdt_skel))); | |
757 | fdt = g_malloc0(size); | |
758 | _FDT((fdt_open_into(fdt_skel, fdt, size))); | |
759 | g_free(fdt_skel); | |
760 | ||
761 | /* Fixup cpu nodes */ | |
5b120785 | 762 | _FDT((spapr_fixup_cpu_dt(fdt, spapr))); |
03d196b7 | 763 | |
6787d27b MR |
764 | if (spapr_dt_cas_updates(spapr, fdt, ov5_updates)) { |
765 | return -1; | |
03d196b7 BR |
766 | } |
767 | ||
768 | /* Pack resulting tree */ | |
769 | _FDT((fdt_pack(fdt))); | |
770 | ||
771 | if (fdt_totalsize(fdt) + sizeof(hdr) > size) { | |
772 | trace_spapr_cas_failed(size); | |
773 | return -1; | |
774 | } | |
775 | ||
776 | cpu_physical_memory_write(addr, &hdr, sizeof(hdr)); | |
777 | cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt)); | |
778 | trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr)); | |
779 | g_free(fdt); | |
780 | ||
781 | return 0; | |
782 | } | |
783 | ||
3f5dabce DG |
784 | static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt) |
785 | { | |
786 | int rtas; | |
787 | GString *hypertas = g_string_sized_new(256); | |
788 | GString *qemu_hypertas = g_string_sized_new(256); | |
789 | uint32_t refpoints[] = { cpu_to_be32(0x4), cpu_to_be32(0x4) }; | |
790 | uint64_t max_hotplug_addr = spapr->hotplug_memory.base + | |
791 | memory_region_size(&spapr->hotplug_memory.mr); | |
792 | uint32_t lrdr_capacity[] = { | |
793 | cpu_to_be32(max_hotplug_addr >> 32), | |
794 | cpu_to_be32(max_hotplug_addr & 0xffffffff), | |
795 | 0, cpu_to_be32(SPAPR_MEMORY_BLOCK_SIZE), | |
796 | cpu_to_be32(max_cpus / smp_threads), | |
797 | }; | |
798 | ||
799 | _FDT(rtas = fdt_add_subnode(fdt, 0, "rtas")); | |
800 | ||
801 | /* hypertas */ | |
802 | add_str(hypertas, "hcall-pft"); | |
803 | add_str(hypertas, "hcall-term"); | |
804 | add_str(hypertas, "hcall-dabr"); | |
805 | add_str(hypertas, "hcall-interrupt"); | |
806 | add_str(hypertas, "hcall-tce"); | |
807 | add_str(hypertas, "hcall-vio"); | |
808 | add_str(hypertas, "hcall-splpar"); | |
809 | add_str(hypertas, "hcall-bulk"); | |
810 | add_str(hypertas, "hcall-set-mode"); | |
811 | add_str(hypertas, "hcall-sprg0"); | |
812 | add_str(hypertas, "hcall-copy"); | |
813 | add_str(hypertas, "hcall-debug"); | |
814 | add_str(qemu_hypertas, "hcall-memop1"); | |
815 | ||
816 | if (!kvm_enabled() || kvmppc_spapr_use_multitce()) { | |
817 | add_str(hypertas, "hcall-multi-tce"); | |
818 | } | |
819 | _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions", | |
820 | hypertas->str, hypertas->len)); | |
821 | g_string_free(hypertas, TRUE); | |
822 | _FDT(fdt_setprop(fdt, rtas, "qemu,hypertas-functions", | |
823 | qemu_hypertas->str, qemu_hypertas->len)); | |
824 | g_string_free(qemu_hypertas, TRUE); | |
825 | ||
826 | _FDT(fdt_setprop(fdt, rtas, "ibm,associativity-reference-points", | |
827 | refpoints, sizeof(refpoints))); | |
828 | ||
829 | _FDT(fdt_setprop_cell(fdt, rtas, "rtas-error-log-max", | |
830 | RTAS_ERROR_LOG_MAX)); | |
831 | _FDT(fdt_setprop_cell(fdt, rtas, "rtas-event-scan-rate", | |
832 | RTAS_EVENT_SCAN_RATE)); | |
833 | ||
834 | if (msi_nonbroken) { | |
835 | _FDT(fdt_setprop(fdt, rtas, "ibm,change-msix-capable", NULL, 0)); | |
836 | } | |
837 | ||
838 | /* | |
839 | * According to PAPR, rtas ibm,os-term does not guarantee a return | |
840 | * back to the guest cpu. | |
841 | * | |
842 | * While an additional ibm,extended-os-term property indicates | |
843 | * that rtas call return will always occur. Set this property. | |
844 | */ | |
845 | _FDT(fdt_setprop(fdt, rtas, "ibm,extended-os-term", NULL, 0)); | |
846 | ||
847 | _FDT(fdt_setprop(fdt, rtas, "ibm,lrdr-capacity", | |
848 | lrdr_capacity, sizeof(lrdr_capacity))); | |
849 | ||
850 | spapr_dt_rtas_tokens(fdt, rtas); | |
851 | } | |
852 | ||
9fb4541f SB |
853 | /* Prepare ibm,arch-vec-5-platform-support, which indicates the MMU features |
854 | * that the guest may request and thus the valid values for bytes 24..26 of | |
855 | * option vector 5: */ | |
856 | static void spapr_dt_ov5_platform_support(void *fdt, int chosen) | |
857 | { | |
545d6e2b SJS |
858 | PowerPCCPU *first_ppc_cpu = POWERPC_CPU(first_cpu); |
859 | ||
9fb4541f SB |
860 | char val[2 * 3] = { |
861 | 24, 0x00, /* Hash/Radix, filled in below. */ | |
862 | 25, 0x00, /* Hash options: Segment Tables == no, GTSE == no. */ | |
863 | 26, 0x40, /* Radix options: GTSE == yes. */ | |
864 | }; | |
865 | ||
866 | if (kvm_enabled()) { | |
867 | if (kvmppc_has_cap_mmu_radix() && kvmppc_has_cap_mmu_hash_v3()) { | |
868 | val[1] = 0x80; /* OV5_MMU_BOTH */ | |
869 | } else if (kvmppc_has_cap_mmu_radix()) { | |
870 | val[1] = 0x40; /* OV5_MMU_RADIX_300 */ | |
871 | } else { | |
872 | val[1] = 0x00; /* Hash */ | |
873 | } | |
874 | } else { | |
545d6e2b SJS |
875 | if (first_ppc_cpu->env.mmu_model & POWERPC_MMU_V3) { |
876 | /* V3 MMU supports both hash and radix (with dynamic switching) */ | |
877 | val[1] = 0xC0; | |
878 | } else { | |
879 | /* Otherwise we can only do hash */ | |
880 | val[1] = 0x00; | |
881 | } | |
9fb4541f SB |
882 | } |
883 | _FDT(fdt_setprop(fdt, chosen, "ibm,arch-vec-5-platform-support", | |
884 | val, sizeof(val))); | |
885 | } | |
886 | ||
7c866c6a DG |
887 | static void spapr_dt_chosen(sPAPRMachineState *spapr, void *fdt) |
888 | { | |
889 | MachineState *machine = MACHINE(spapr); | |
890 | int chosen; | |
891 | const char *boot_device = machine->boot_order; | |
892 | char *stdout_path = spapr_vio_stdout_path(spapr->vio_bus); | |
893 | size_t cb = 0; | |
894 | char *bootlist = get_boot_devices_list(&cb, true); | |
7c866c6a DG |
895 | |
896 | _FDT(chosen = fdt_add_subnode(fdt, 0, "chosen")); | |
897 | ||
7c866c6a DG |
898 | _FDT(fdt_setprop_string(fdt, chosen, "bootargs", machine->kernel_cmdline)); |
899 | _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-start", | |
900 | spapr->initrd_base)); | |
901 | _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-end", | |
902 | spapr->initrd_base + spapr->initrd_size)); | |
903 | ||
904 | if (spapr->kernel_size) { | |
905 | uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR), | |
906 | cpu_to_be64(spapr->kernel_size) }; | |
907 | ||
908 | _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel", | |
909 | &kprop, sizeof(kprop))); | |
910 | if (spapr->kernel_le) { | |
911 | _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel-le", NULL, 0)); | |
912 | } | |
913 | } | |
914 | if (boot_menu) { | |
915 | _FDT((fdt_setprop_cell(fdt, chosen, "qemu,boot-menu", boot_menu))); | |
916 | } | |
917 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-width", graphic_width)); | |
918 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-height", graphic_height)); | |
919 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-depth", graphic_depth)); | |
920 | ||
921 | if (cb && bootlist) { | |
922 | int i; | |
923 | ||
924 | for (i = 0; i < cb; i++) { | |
925 | if (bootlist[i] == '\n') { | |
926 | bootlist[i] = ' '; | |
927 | } | |
928 | } | |
929 | _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-list", bootlist)); | |
930 | } | |
931 | ||
932 | if (boot_device && strlen(boot_device)) { | |
933 | _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-device", boot_device)); | |
934 | } | |
935 | ||
936 | if (!spapr->has_graphics && stdout_path) { | |
937 | _FDT(fdt_setprop_string(fdt, chosen, "linux,stdout-path", stdout_path)); | |
938 | } | |
939 | ||
9fb4541f SB |
940 | spapr_dt_ov5_platform_support(fdt, chosen); |
941 | ||
7c866c6a DG |
942 | g_free(stdout_path); |
943 | g_free(bootlist); | |
944 | } | |
945 | ||
fca5f2dc DG |
946 | static void spapr_dt_hypervisor(sPAPRMachineState *spapr, void *fdt) |
947 | { | |
948 | /* The /hypervisor node isn't in PAPR - this is a hack to allow PR | |
949 | * KVM to work under pHyp with some guest co-operation */ | |
950 | int hypervisor; | |
951 | uint8_t hypercall[16]; | |
952 | ||
953 | _FDT(hypervisor = fdt_add_subnode(fdt, 0, "hypervisor")); | |
954 | /* indicate KVM hypercall interface */ | |
955 | _FDT(fdt_setprop_string(fdt, hypervisor, "compatible", "linux,kvm")); | |
956 | if (kvmppc_has_cap_fixup_hcalls()) { | |
957 | /* | |
958 | * Older KVM versions with older guest kernels were broken | |
959 | * with the magic page, don't allow the guest to map it. | |
960 | */ | |
961 | if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall, | |
962 | sizeof(hypercall))) { | |
963 | _FDT(fdt_setprop(fdt, hypervisor, "hcall-instructions", | |
964 | hypercall, sizeof(hypercall))); | |
965 | } | |
966 | } | |
967 | } | |
968 | ||
997b6cfc DG |
969 | static void *spapr_build_fdt(sPAPRMachineState *spapr, |
970 | hwaddr rtas_addr, | |
971 | hwaddr rtas_size) | |
a3467baa | 972 | { |
5b2128d2 | 973 | MachineState *machine = MACHINE(qdev_get_machine()); |
3c0c47e3 | 974 | MachineClass *mc = MACHINE_GET_CLASS(machine); |
c20d332a | 975 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); |
7c866c6a | 976 | int ret; |
a3467baa | 977 | void *fdt; |
3384f95c | 978 | sPAPRPHBState *phb; |
398a0bd5 | 979 | char *buf; |
71cd4dac | 980 | int smt = kvmppc_smt_threads(); |
a3467baa | 981 | |
398a0bd5 DG |
982 | fdt = g_malloc0(FDT_MAX_SIZE); |
983 | _FDT((fdt_create_empty_tree(fdt, FDT_MAX_SIZE))); | |
a3467baa | 984 | |
398a0bd5 DG |
985 | /* Root node */ |
986 | _FDT(fdt_setprop_string(fdt, 0, "device_type", "chrp")); | |
987 | _FDT(fdt_setprop_string(fdt, 0, "model", "IBM pSeries (emulated by qemu)")); | |
988 | _FDT(fdt_setprop_string(fdt, 0, "compatible", "qemu,pseries")); | |
989 | ||
990 | /* | |
991 | * Add info to guest to indentify which host is it being run on | |
992 | * and what is the uuid of the guest | |
993 | */ | |
994 | if (kvmppc_get_host_model(&buf)) { | |
995 | _FDT(fdt_setprop_string(fdt, 0, "host-model", buf)); | |
996 | g_free(buf); | |
997 | } | |
998 | if (kvmppc_get_host_serial(&buf)) { | |
999 | _FDT(fdt_setprop_string(fdt, 0, "host-serial", buf)); | |
1000 | g_free(buf); | |
1001 | } | |
1002 | ||
1003 | buf = qemu_uuid_unparse_strdup(&qemu_uuid); | |
1004 | ||
1005 | _FDT(fdt_setprop_string(fdt, 0, "vm,uuid", buf)); | |
1006 | if (qemu_uuid_set) { | |
1007 | _FDT(fdt_setprop_string(fdt, 0, "system-id", buf)); | |
1008 | } | |
1009 | g_free(buf); | |
1010 | ||
1011 | if (qemu_get_vm_name()) { | |
1012 | _FDT(fdt_setprop_string(fdt, 0, "ibm,partition-name", | |
1013 | qemu_get_vm_name())); | |
1014 | } | |
1015 | ||
1016 | _FDT(fdt_setprop_cell(fdt, 0, "#address-cells", 2)); | |
1017 | _FDT(fdt_setprop_cell(fdt, 0, "#size-cells", 2)); | |
4040ab72 | 1018 | |
9b9a1908 | 1019 | /* /interrupt controller */ |
71cd4dac | 1020 | spapr_dt_xics(DIV_ROUND_UP(max_cpus * smt, smp_threads), fdt, PHANDLE_XICP); |
9b9a1908 | 1021 | |
e8f986fc BR |
1022 | ret = spapr_populate_memory(spapr, fdt); |
1023 | if (ret < 0) { | |
ce9863b7 | 1024 | error_report("couldn't setup memory nodes in fdt"); |
e8f986fc | 1025 | exit(1); |
7f763a5d DG |
1026 | } |
1027 | ||
bf5a6696 DG |
1028 | /* /vdevice */ |
1029 | spapr_dt_vdevice(spapr->vio_bus, fdt); | |
4040ab72 | 1030 | |
4d9392be TH |
1031 | if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) { |
1032 | ret = spapr_rng_populate_dt(fdt); | |
1033 | if (ret < 0) { | |
ce9863b7 | 1034 | error_report("could not set up rng device in the fdt"); |
4d9392be TH |
1035 | exit(1); |
1036 | } | |
1037 | } | |
1038 | ||
3384f95c | 1039 | QLIST_FOREACH(phb, &spapr->phbs, list) { |
e0fdbd7c | 1040 | ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt); |
da34fed7 TH |
1041 | if (ret < 0) { |
1042 | error_report("couldn't setup PCI devices in fdt"); | |
1043 | exit(1); | |
1044 | } | |
3384f95c DG |
1045 | } |
1046 | ||
0da6f3fe BR |
1047 | /* cpus */ |
1048 | spapr_populate_cpus_dt_node(fdt, spapr); | |
6e806cc3 | 1049 | |
c20d332a BR |
1050 | if (smc->dr_lmb_enabled) { |
1051 | _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB)); | |
1052 | } | |
1053 | ||
c5514d0e | 1054 | if (mc->has_hotpluggable_cpus) { |
af81cf32 BR |
1055 | int offset = fdt_path_offset(fdt, "/cpus"); |
1056 | ret = spapr_drc_populate_dt(fdt, offset, NULL, | |
1057 | SPAPR_DR_CONNECTOR_TYPE_CPU); | |
1058 | if (ret < 0) { | |
1059 | error_report("Couldn't set up CPU DR device tree properties"); | |
1060 | exit(1); | |
1061 | } | |
1062 | } | |
1063 | ||
ffb1e275 | 1064 | /* /event-sources */ |
ffbb1705 | 1065 | spapr_dt_events(spapr, fdt); |
ffb1e275 | 1066 | |
3f5dabce DG |
1067 | /* /rtas */ |
1068 | spapr_dt_rtas(spapr, fdt); | |
1069 | ||
7c866c6a DG |
1070 | /* /chosen */ |
1071 | spapr_dt_chosen(spapr, fdt); | |
cf6e5223 | 1072 | |
fca5f2dc DG |
1073 | /* /hypervisor */ |
1074 | if (kvm_enabled()) { | |
1075 | spapr_dt_hypervisor(spapr, fdt); | |
1076 | } | |
1077 | ||
cf6e5223 DG |
1078 | /* Build memory reserve map */ |
1079 | if (spapr->kernel_size) { | |
1080 | _FDT((fdt_add_mem_rsv(fdt, KERNEL_LOAD_ADDR, spapr->kernel_size))); | |
1081 | } | |
1082 | if (spapr->initrd_size) { | |
1083 | _FDT((fdt_add_mem_rsv(fdt, spapr->initrd_base, spapr->initrd_size))); | |
1084 | } | |
1085 | ||
6787d27b MR |
1086 | /* ibm,client-architecture-support updates */ |
1087 | ret = spapr_dt_cas_updates(spapr, fdt, spapr->ov5_cas); | |
1088 | if (ret < 0) { | |
1089 | error_report("couldn't setup CAS properties fdt"); | |
1090 | exit(1); | |
1091 | } | |
1092 | ||
997b6cfc | 1093 | return fdt; |
9fdf0c29 DG |
1094 | } |
1095 | ||
1096 | static uint64_t translate_kernel_address(void *opaque, uint64_t addr) | |
1097 | { | |
1098 | return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR; | |
1099 | } | |
1100 | ||
1d1be34d DG |
1101 | static void emulate_spapr_hypercall(PPCVirtualHypervisor *vhyp, |
1102 | PowerPCCPU *cpu) | |
9fdf0c29 | 1103 | { |
1b14670a AF |
1104 | CPUPPCState *env = &cpu->env; |
1105 | ||
8d04fb55 JK |
1106 | /* The TCG path should also be holding the BQL at this point */ |
1107 | g_assert(qemu_mutex_iothread_locked()); | |
1108 | ||
efcb9383 DG |
1109 | if (msr_pr) { |
1110 | hcall_dprintf("Hypercall made with MSR[PR]=1\n"); | |
1111 | env->gpr[3] = H_PRIVILEGE; | |
1112 | } else { | |
aa100fa4 | 1113 | env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]); |
efcb9383 | 1114 | } |
9fdf0c29 DG |
1115 | } |
1116 | ||
9861bb3e SJS |
1117 | static uint64_t spapr_get_patbe(PPCVirtualHypervisor *vhyp) |
1118 | { | |
1119 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1120 | ||
1121 | return spapr->patb_entry; | |
1122 | } | |
1123 | ||
e6b8fd24 SMJ |
1124 | #define HPTE(_table, _i) (void *)(((uint64_t *)(_table)) + ((_i) * 2)) |
1125 | #define HPTE_VALID(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID) | |
1126 | #define HPTE_DIRTY(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY) | |
1127 | #define CLEAN_HPTE(_hpte) ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY)) | |
1128 | #define DIRTY_HPTE(_hpte) ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY)) | |
1129 | ||
715c5407 DG |
1130 | /* |
1131 | * Get the fd to access the kernel htab, re-opening it if necessary | |
1132 | */ | |
1133 | static int get_htab_fd(sPAPRMachineState *spapr) | |
1134 | { | |
1135 | if (spapr->htab_fd >= 0) { | |
1136 | return spapr->htab_fd; | |
1137 | } | |
1138 | ||
1139 | spapr->htab_fd = kvmppc_get_htab_fd(false); | |
1140 | if (spapr->htab_fd < 0) { | |
1141 | error_report("Unable to open fd for reading hash table from KVM: %s", | |
1142 | strerror(errno)); | |
1143 | } | |
1144 | ||
1145 | return spapr->htab_fd; | |
1146 | } | |
1147 | ||
b4db5413 | 1148 | void close_htab_fd(sPAPRMachineState *spapr) |
715c5407 DG |
1149 | { |
1150 | if (spapr->htab_fd >= 0) { | |
1151 | close(spapr->htab_fd); | |
1152 | } | |
1153 | spapr->htab_fd = -1; | |
1154 | } | |
1155 | ||
e57ca75c DG |
1156 | static hwaddr spapr_hpt_mask(PPCVirtualHypervisor *vhyp) |
1157 | { | |
1158 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1159 | ||
1160 | return HTAB_SIZE(spapr) / HASH_PTEG_SIZE_64 - 1; | |
1161 | } | |
1162 | ||
1163 | static const ppc_hash_pte64_t *spapr_map_hptes(PPCVirtualHypervisor *vhyp, | |
1164 | hwaddr ptex, int n) | |
1165 | { | |
1166 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1167 | hwaddr pte_offset = ptex * HASH_PTE_SIZE_64; | |
1168 | ||
1169 | if (!spapr->htab) { | |
1170 | /* | |
1171 | * HTAB is controlled by KVM. Fetch into temporary buffer | |
1172 | */ | |
1173 | ppc_hash_pte64_t *hptes = g_malloc(n * HASH_PTE_SIZE_64); | |
1174 | kvmppc_read_hptes(hptes, ptex, n); | |
1175 | return hptes; | |
1176 | } | |
1177 | ||
1178 | /* | |
1179 | * HTAB is controlled by QEMU. Just point to the internally | |
1180 | * accessible PTEG. | |
1181 | */ | |
1182 | return (const ppc_hash_pte64_t *)(spapr->htab + pte_offset); | |
1183 | } | |
1184 | ||
1185 | static void spapr_unmap_hptes(PPCVirtualHypervisor *vhyp, | |
1186 | const ppc_hash_pte64_t *hptes, | |
1187 | hwaddr ptex, int n) | |
1188 | { | |
1189 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1190 | ||
1191 | if (!spapr->htab) { | |
1192 | g_free((void *)hptes); | |
1193 | } | |
1194 | ||
1195 | /* Nothing to do for qemu managed HPT */ | |
1196 | } | |
1197 | ||
1198 | static void spapr_store_hpte(PPCVirtualHypervisor *vhyp, hwaddr ptex, | |
1199 | uint64_t pte0, uint64_t pte1) | |
1200 | { | |
1201 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1202 | hwaddr offset = ptex * HASH_PTE_SIZE_64; | |
1203 | ||
1204 | if (!spapr->htab) { | |
1205 | kvmppc_write_hpte(ptex, pte0, pte1); | |
1206 | } else { | |
1207 | stq_p(spapr->htab + offset, pte0); | |
1208 | stq_p(spapr->htab + offset + HASH_PTE_SIZE_64 / 2, pte1); | |
1209 | } | |
1210 | } | |
1211 | ||
8dfe8e7f DG |
1212 | static int spapr_hpt_shift_for_ramsize(uint64_t ramsize) |
1213 | { | |
1214 | int shift; | |
1215 | ||
1216 | /* We aim for a hash table of size 1/128 the size of RAM (rounded | |
1217 | * up). The PAPR recommendation is actually 1/64 of RAM size, but | |
1218 | * that's much more than is needed for Linux guests */ | |
1219 | shift = ctz64(pow2ceil(ramsize)) - 7; | |
1220 | shift = MAX(shift, 18); /* Minimum architected size */ | |
1221 | shift = MIN(shift, 46); /* Maximum architected size */ | |
1222 | return shift; | |
1223 | } | |
1224 | ||
06ec79e8 BR |
1225 | void spapr_free_hpt(sPAPRMachineState *spapr) |
1226 | { | |
1227 | g_free(spapr->htab); | |
1228 | spapr->htab = NULL; | |
1229 | spapr->htab_shift = 0; | |
1230 | close_htab_fd(spapr); | |
1231 | } | |
1232 | ||
c5f54f3e DG |
1233 | static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift, |
1234 | Error **errp) | |
7f763a5d | 1235 | { |
c5f54f3e DG |
1236 | long rc; |
1237 | ||
1238 | /* Clean up any HPT info from a previous boot */ | |
06ec79e8 | 1239 | spapr_free_hpt(spapr); |
c5f54f3e DG |
1240 | |
1241 | rc = kvmppc_reset_htab(shift); | |
1242 | if (rc < 0) { | |
1243 | /* kernel-side HPT needed, but couldn't allocate one */ | |
1244 | error_setg_errno(errp, errno, | |
1245 | "Failed to allocate KVM HPT of order %d (try smaller maxmem?)", | |
1246 | shift); | |
1247 | /* This is almost certainly fatal, but if the caller really | |
1248 | * wants to carry on with shift == 0, it's welcome to try */ | |
1249 | } else if (rc > 0) { | |
1250 | /* kernel-side HPT allocated */ | |
1251 | if (rc != shift) { | |
1252 | error_setg(errp, | |
1253 | "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)", | |
1254 | shift, rc); | |
7735feda BR |
1255 | } |
1256 | ||
7f763a5d | 1257 | spapr->htab_shift = shift; |
c18ad9a5 | 1258 | spapr->htab = NULL; |
b817772a | 1259 | } else { |
c5f54f3e DG |
1260 | /* kernel-side HPT not needed, allocate in userspace instead */ |
1261 | size_t size = 1ULL << shift; | |
1262 | int i; | |
b817772a | 1263 | |
c5f54f3e DG |
1264 | spapr->htab = qemu_memalign(size, size); |
1265 | if (!spapr->htab) { | |
1266 | error_setg_errno(errp, errno, | |
1267 | "Could not allocate HPT of order %d", shift); | |
1268 | return; | |
7735feda BR |
1269 | } |
1270 | ||
c5f54f3e DG |
1271 | memset(spapr->htab, 0, size); |
1272 | spapr->htab_shift = shift; | |
e6b8fd24 | 1273 | |
c5f54f3e DG |
1274 | for (i = 0; i < size / HASH_PTE_SIZE_64; i++) { |
1275 | DIRTY_HPTE(HPTE(spapr->htab, i)); | |
e6b8fd24 | 1276 | } |
7f763a5d | 1277 | } |
9fdf0c29 DG |
1278 | } |
1279 | ||
b4db5413 SJS |
1280 | void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr) |
1281 | { | |
1282 | spapr_reallocate_hpt(spapr, | |
1283 | spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size), | |
1284 | &error_fatal); | |
1285 | if (spapr->vrma_adjust) { | |
1286 | spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), | |
1287 | spapr->htab_shift); | |
1288 | } | |
1289 | /* We're setting up a hash table, so that means we're not radix */ | |
1290 | spapr->patb_entry = 0; | |
1291 | } | |
1292 | ||
4f01a637 | 1293 | static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque) |
9e3f9733 AG |
1294 | { |
1295 | bool matched = false; | |
1296 | ||
1297 | if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) { | |
1298 | matched = true; | |
1299 | } | |
1300 | ||
1301 | if (!matched) { | |
1302 | error_report("Device %s is not supported by this machine yet.", | |
1303 | qdev_fw_name(DEVICE(sbdev))); | |
1304 | exit(1); | |
1305 | } | |
9e3f9733 AG |
1306 | } |
1307 | ||
c8787ad4 | 1308 | static void ppc_spapr_reset(void) |
a3467baa | 1309 | { |
c5f54f3e DG |
1310 | MachineState *machine = MACHINE(qdev_get_machine()); |
1311 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); | |
182735ef | 1312 | PowerPCCPU *first_ppc_cpu; |
b7d1f77a | 1313 | uint32_t rtas_limit; |
cae172ab | 1314 | hwaddr rtas_addr, fdt_addr; |
997b6cfc DG |
1315 | void *fdt; |
1316 | int rc; | |
259186a7 | 1317 | |
9e3f9733 AG |
1318 | /* Check for unknown sysbus devices */ |
1319 | foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL); | |
1320 | ||
b4db5413 SJS |
1321 | if (kvm_enabled() && kvmppc_has_cap_mmu_radix()) { |
1322 | /* If using KVM with radix mode available, VCPUs can be started | |
1323 | * without a HPT because KVM will start them in radix mode. | |
1324 | * Set the GR bit in PATB so that we know there is no HPT. */ | |
1325 | spapr->patb_entry = PATBE1_GR; | |
1326 | } else { | |
1327 | spapr->patb_entry = 0; | |
1328 | spapr_setup_hpt_and_vrma(spapr); | |
c5f54f3e | 1329 | } |
a3467baa | 1330 | |
c8787ad4 | 1331 | qemu_devices_reset(); |
a3467baa | 1332 | |
b7d1f77a BH |
1333 | /* |
1334 | * We place the device tree and RTAS just below either the top of the RMA, | |
1335 | * or just below 2GB, whichever is lowere, so that it can be | |
1336 | * processed with 32-bit real mode code if necessary | |
1337 | */ | |
1338 | rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR); | |
cae172ab DG |
1339 | rtas_addr = rtas_limit - RTAS_MAX_SIZE; |
1340 | fdt_addr = rtas_addr - FDT_MAX_SIZE; | |
b7d1f77a | 1341 | |
6787d27b MR |
1342 | /* if this reset wasn't generated by CAS, we should reset our |
1343 | * negotiated options and start from scratch */ | |
1344 | if (!spapr->cas_reboot) { | |
1345 | spapr_ovec_cleanup(spapr->ov5_cas); | |
1346 | spapr->ov5_cas = spapr_ovec_new(); | |
1347 | } | |
1348 | ||
cae172ab | 1349 | fdt = spapr_build_fdt(spapr, rtas_addr, spapr->rtas_size); |
a3467baa | 1350 | |
2cac78c1 | 1351 | spapr_load_rtas(spapr, fdt, rtas_addr); |
b7d1f77a | 1352 | |
997b6cfc DG |
1353 | rc = fdt_pack(fdt); |
1354 | ||
1355 | /* Should only fail if we've built a corrupted tree */ | |
1356 | assert(rc == 0); | |
1357 | ||
1358 | if (fdt_totalsize(fdt) > FDT_MAX_SIZE) { | |
1359 | error_report("FDT too big ! 0x%x bytes (max is 0x%x)", | |
1360 | fdt_totalsize(fdt), FDT_MAX_SIZE); | |
1361 | exit(1); | |
1362 | } | |
1363 | ||
1364 | /* Load the fdt */ | |
1365 | qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt)); | |
cae172ab | 1366 | cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt)); |
997b6cfc DG |
1367 | g_free(fdt); |
1368 | ||
a3467baa | 1369 | /* Set up the entry state */ |
182735ef | 1370 | first_ppc_cpu = POWERPC_CPU(first_cpu); |
cae172ab | 1371 | first_ppc_cpu->env.gpr[3] = fdt_addr; |
182735ef AF |
1372 | first_ppc_cpu->env.gpr[5] = 0; |
1373 | first_cpu->halted = 0; | |
1b718907 | 1374 | first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT; |
a3467baa | 1375 | |
6787d27b | 1376 | spapr->cas_reboot = false; |
a3467baa DG |
1377 | } |
1378 | ||
28e02042 | 1379 | static void spapr_create_nvram(sPAPRMachineState *spapr) |
639e8102 | 1380 | { |
2ff3de68 | 1381 | DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram"); |
3978b863 | 1382 | DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0); |
639e8102 | 1383 | |
3978b863 | 1384 | if (dinfo) { |
6231a6da MA |
1385 | qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo), |
1386 | &error_fatal); | |
639e8102 DG |
1387 | } |
1388 | ||
1389 | qdev_init_nofail(dev); | |
1390 | ||
1391 | spapr->nvram = (struct sPAPRNVRAM *)dev; | |
1392 | } | |
1393 | ||
28e02042 | 1394 | static void spapr_rtc_create(sPAPRMachineState *spapr) |
28df36a1 | 1395 | { |
147ff807 CLG |
1396 | object_initialize(&spapr->rtc, sizeof(spapr->rtc), TYPE_SPAPR_RTC); |
1397 | object_property_add_child(OBJECT(spapr), "rtc", OBJECT(&spapr->rtc), | |
1398 | &error_fatal); | |
1399 | object_property_set_bool(OBJECT(&spapr->rtc), true, "realized", | |
1400 | &error_fatal); | |
1401 | object_property_add_alias(OBJECT(spapr), "rtc-time", OBJECT(&spapr->rtc), | |
1402 | "date", &error_fatal); | |
28df36a1 DG |
1403 | } |
1404 | ||
8c57b867 | 1405 | /* Returns whether we want to use VGA or not */ |
14c6a894 | 1406 | static bool spapr_vga_init(PCIBus *pci_bus, Error **errp) |
f28359d8 | 1407 | { |
8c57b867 | 1408 | switch (vga_interface_type) { |
8c57b867 | 1409 | case VGA_NONE: |
7effdaa3 MW |
1410 | return false; |
1411 | case VGA_DEVICE: | |
1412 | return true; | |
1ddcae82 | 1413 | case VGA_STD: |
b798c190 | 1414 | case VGA_VIRTIO: |
1ddcae82 | 1415 | return pci_vga_init(pci_bus) != NULL; |
8c57b867 | 1416 | default: |
14c6a894 DG |
1417 | error_setg(errp, |
1418 | "Unsupported VGA mode, only -vga std or -vga virtio is supported"); | |
1419 | return false; | |
f28359d8 | 1420 | } |
f28359d8 LZ |
1421 | } |
1422 | ||
880ae7de DG |
1423 | static int spapr_post_load(void *opaque, int version_id) |
1424 | { | |
28e02042 | 1425 | sPAPRMachineState *spapr = (sPAPRMachineState *)opaque; |
880ae7de DG |
1426 | int err = 0; |
1427 | ||
a7ff1212 | 1428 | if (!object_dynamic_cast(OBJECT(spapr->ics), TYPE_ICS_KVM)) { |
5bc8d26d CLG |
1429 | CPUState *cs; |
1430 | CPU_FOREACH(cs) { | |
1431 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
1432 | icp_resend(ICP(cpu->intc)); | |
a7ff1212 CLG |
1433 | } |
1434 | } | |
1435 | ||
631b22ea | 1436 | /* In earlier versions, there was no separate qdev for the PAPR |
880ae7de DG |
1437 | * RTC, so the RTC offset was stored directly in sPAPREnvironment. |
1438 | * So when migrating from those versions, poke the incoming offset | |
1439 | * value into the RTC device */ | |
1440 | if (version_id < 3) { | |
147ff807 | 1441 | err = spapr_rtc_import_offset(&spapr->rtc, spapr->rtc_offset); |
880ae7de DG |
1442 | } |
1443 | ||
1444 | return err; | |
1445 | } | |
1446 | ||
1447 | static bool version_before_3(void *opaque, int version_id) | |
1448 | { | |
1449 | return version_id < 3; | |
1450 | } | |
1451 | ||
62ef3760 MR |
1452 | static bool spapr_ov5_cas_needed(void *opaque) |
1453 | { | |
1454 | sPAPRMachineState *spapr = opaque; | |
1455 | sPAPROptionVector *ov5_mask = spapr_ovec_new(); | |
1456 | sPAPROptionVector *ov5_legacy = spapr_ovec_new(); | |
1457 | sPAPROptionVector *ov5_removed = spapr_ovec_new(); | |
1458 | bool cas_needed; | |
1459 | ||
1460 | /* Prior to the introduction of sPAPROptionVector, we had two option | |
1461 | * vectors we dealt with: OV5_FORM1_AFFINITY, and OV5_DRCONF_MEMORY. | |
1462 | * Both of these options encode machine topology into the device-tree | |
1463 | * in such a way that the now-booted OS should still be able to interact | |
1464 | * appropriately with QEMU regardless of what options were actually | |
1465 | * negotiatied on the source side. | |
1466 | * | |
1467 | * As such, we can avoid migrating the CAS-negotiated options if these | |
1468 | * are the only options available on the current machine/platform. | |
1469 | * Since these are the only options available for pseries-2.7 and | |
1470 | * earlier, this allows us to maintain old->new/new->old migration | |
1471 | * compatibility. | |
1472 | * | |
1473 | * For QEMU 2.8+, there are additional CAS-negotiatable options available | |
1474 | * via default pseries-2.8 machines and explicit command-line parameters. | |
1475 | * Some of these options, like OV5_HP_EVT, *do* require QEMU to be aware | |
1476 | * of the actual CAS-negotiated values to continue working properly. For | |
1477 | * example, availability of memory unplug depends on knowing whether | |
1478 | * OV5_HP_EVT was negotiated via CAS. | |
1479 | * | |
1480 | * Thus, for any cases where the set of available CAS-negotiatable | |
1481 | * options extends beyond OV5_FORM1_AFFINITY and OV5_DRCONF_MEMORY, we | |
1482 | * include the CAS-negotiated options in the migration stream. | |
1483 | */ | |
1484 | spapr_ovec_set(ov5_mask, OV5_FORM1_AFFINITY); | |
1485 | spapr_ovec_set(ov5_mask, OV5_DRCONF_MEMORY); | |
1486 | ||
1487 | /* spapr_ovec_diff returns true if bits were removed. we avoid using | |
1488 | * the mask itself since in the future it's possible "legacy" bits may be | |
1489 | * removed via machine options, which could generate a false positive | |
1490 | * that breaks migration. | |
1491 | */ | |
1492 | spapr_ovec_intersect(ov5_legacy, spapr->ov5, ov5_mask); | |
1493 | cas_needed = spapr_ovec_diff(ov5_removed, spapr->ov5, ov5_legacy); | |
1494 | ||
1495 | spapr_ovec_cleanup(ov5_mask); | |
1496 | spapr_ovec_cleanup(ov5_legacy); | |
1497 | spapr_ovec_cleanup(ov5_removed); | |
1498 | ||
1499 | return cas_needed; | |
1500 | } | |
1501 | ||
1502 | static const VMStateDescription vmstate_spapr_ov5_cas = { | |
1503 | .name = "spapr_option_vector_ov5_cas", | |
1504 | .version_id = 1, | |
1505 | .minimum_version_id = 1, | |
1506 | .needed = spapr_ov5_cas_needed, | |
1507 | .fields = (VMStateField[]) { | |
1508 | VMSTATE_STRUCT_POINTER_V(ov5_cas, sPAPRMachineState, 1, | |
1509 | vmstate_spapr_ovec, sPAPROptionVector), | |
1510 | VMSTATE_END_OF_LIST() | |
1511 | }, | |
1512 | }; | |
1513 | ||
9861bb3e SJS |
1514 | static bool spapr_patb_entry_needed(void *opaque) |
1515 | { | |
1516 | sPAPRMachineState *spapr = opaque; | |
1517 | ||
1518 | return !!spapr->patb_entry; | |
1519 | } | |
1520 | ||
1521 | static const VMStateDescription vmstate_spapr_patb_entry = { | |
1522 | .name = "spapr_patb_entry", | |
1523 | .version_id = 1, | |
1524 | .minimum_version_id = 1, | |
1525 | .needed = spapr_patb_entry_needed, | |
1526 | .fields = (VMStateField[]) { | |
1527 | VMSTATE_UINT64(patb_entry, sPAPRMachineState), | |
1528 | VMSTATE_END_OF_LIST() | |
1529 | }, | |
1530 | }; | |
1531 | ||
4be21d56 DG |
1532 | static const VMStateDescription vmstate_spapr = { |
1533 | .name = "spapr", | |
880ae7de | 1534 | .version_id = 3, |
4be21d56 | 1535 | .minimum_version_id = 1, |
880ae7de | 1536 | .post_load = spapr_post_load, |
3aff6c2f | 1537 | .fields = (VMStateField[]) { |
880ae7de DG |
1538 | /* used to be @next_irq */ |
1539 | VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4), | |
4be21d56 DG |
1540 | |
1541 | /* RTC offset */ | |
28e02042 | 1542 | VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3), |
880ae7de | 1543 | |
28e02042 | 1544 | VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2), |
4be21d56 DG |
1545 | VMSTATE_END_OF_LIST() |
1546 | }, | |
62ef3760 MR |
1547 | .subsections = (const VMStateDescription*[]) { |
1548 | &vmstate_spapr_ov5_cas, | |
9861bb3e | 1549 | &vmstate_spapr_patb_entry, |
62ef3760 MR |
1550 | NULL |
1551 | } | |
4be21d56 DG |
1552 | }; |
1553 | ||
4be21d56 DG |
1554 | static int htab_save_setup(QEMUFile *f, void *opaque) |
1555 | { | |
28e02042 | 1556 | sPAPRMachineState *spapr = opaque; |
4be21d56 | 1557 | |
4be21d56 DG |
1558 | /* "Iteration" header */ |
1559 | qemu_put_be32(f, spapr->htab_shift); | |
1560 | ||
e68cb8b4 AK |
1561 | if (spapr->htab) { |
1562 | spapr->htab_save_index = 0; | |
1563 | spapr->htab_first_pass = true; | |
1564 | } else { | |
1565 | assert(kvm_enabled()); | |
e68cb8b4 AK |
1566 | } |
1567 | ||
1568 | ||
4be21d56 DG |
1569 | return 0; |
1570 | } | |
1571 | ||
28e02042 | 1572 | static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr, |
4be21d56 DG |
1573 | int64_t max_ns) |
1574 | { | |
378bc217 | 1575 | bool has_timeout = max_ns != -1; |
4be21d56 DG |
1576 | int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; |
1577 | int index = spapr->htab_save_index; | |
bc72ad67 | 1578 | int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
4be21d56 DG |
1579 | |
1580 | assert(spapr->htab_first_pass); | |
1581 | ||
1582 | do { | |
1583 | int chunkstart; | |
1584 | ||
1585 | /* Consume invalid HPTEs */ | |
1586 | while ((index < htabslots) | |
1587 | && !HPTE_VALID(HPTE(spapr->htab, index))) { | |
4be21d56 | 1588 | CLEAN_HPTE(HPTE(spapr->htab, index)); |
24ec2863 | 1589 | index++; |
4be21d56 DG |
1590 | } |
1591 | ||
1592 | /* Consume valid HPTEs */ | |
1593 | chunkstart = index; | |
338c25b6 | 1594 | while ((index < htabslots) && (index - chunkstart < USHRT_MAX) |
4be21d56 | 1595 | && HPTE_VALID(HPTE(spapr->htab, index))) { |
4be21d56 | 1596 | CLEAN_HPTE(HPTE(spapr->htab, index)); |
24ec2863 | 1597 | index++; |
4be21d56 DG |
1598 | } |
1599 | ||
1600 | if (index > chunkstart) { | |
1601 | int n_valid = index - chunkstart; | |
1602 | ||
1603 | qemu_put_be32(f, chunkstart); | |
1604 | qemu_put_be16(f, n_valid); | |
1605 | qemu_put_be16(f, 0); | |
1606 | qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), | |
1607 | HASH_PTE_SIZE_64 * n_valid); | |
1608 | ||
378bc217 DG |
1609 | if (has_timeout && |
1610 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { | |
4be21d56 DG |
1611 | break; |
1612 | } | |
1613 | } | |
1614 | } while ((index < htabslots) && !qemu_file_rate_limit(f)); | |
1615 | ||
1616 | if (index >= htabslots) { | |
1617 | assert(index == htabslots); | |
1618 | index = 0; | |
1619 | spapr->htab_first_pass = false; | |
1620 | } | |
1621 | spapr->htab_save_index = index; | |
1622 | } | |
1623 | ||
28e02042 | 1624 | static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr, |
e68cb8b4 | 1625 | int64_t max_ns) |
4be21d56 DG |
1626 | { |
1627 | bool final = max_ns < 0; | |
1628 | int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; | |
1629 | int examined = 0, sent = 0; | |
1630 | int index = spapr->htab_save_index; | |
bc72ad67 | 1631 | int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
4be21d56 DG |
1632 | |
1633 | assert(!spapr->htab_first_pass); | |
1634 | ||
1635 | do { | |
1636 | int chunkstart, invalidstart; | |
1637 | ||
1638 | /* Consume non-dirty HPTEs */ | |
1639 | while ((index < htabslots) | |
1640 | && !HPTE_DIRTY(HPTE(spapr->htab, index))) { | |
1641 | index++; | |
1642 | examined++; | |
1643 | } | |
1644 | ||
1645 | chunkstart = index; | |
1646 | /* Consume valid dirty HPTEs */ | |
338c25b6 | 1647 | while ((index < htabslots) && (index - chunkstart < USHRT_MAX) |
4be21d56 DG |
1648 | && HPTE_DIRTY(HPTE(spapr->htab, index)) |
1649 | && HPTE_VALID(HPTE(spapr->htab, index))) { | |
1650 | CLEAN_HPTE(HPTE(spapr->htab, index)); | |
1651 | index++; | |
1652 | examined++; | |
1653 | } | |
1654 | ||
1655 | invalidstart = index; | |
1656 | /* Consume invalid dirty HPTEs */ | |
338c25b6 | 1657 | while ((index < htabslots) && (index - invalidstart < USHRT_MAX) |
4be21d56 DG |
1658 | && HPTE_DIRTY(HPTE(spapr->htab, index)) |
1659 | && !HPTE_VALID(HPTE(spapr->htab, index))) { | |
1660 | CLEAN_HPTE(HPTE(spapr->htab, index)); | |
1661 | index++; | |
1662 | examined++; | |
1663 | } | |
1664 | ||
1665 | if (index > chunkstart) { | |
1666 | int n_valid = invalidstart - chunkstart; | |
1667 | int n_invalid = index - invalidstart; | |
1668 | ||
1669 | qemu_put_be32(f, chunkstart); | |
1670 | qemu_put_be16(f, n_valid); | |
1671 | qemu_put_be16(f, n_invalid); | |
1672 | qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), | |
1673 | HASH_PTE_SIZE_64 * n_valid); | |
1674 | sent += index - chunkstart; | |
1675 | ||
bc72ad67 | 1676 | if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { |
4be21d56 DG |
1677 | break; |
1678 | } | |
1679 | } | |
1680 | ||
1681 | if (examined >= htabslots) { | |
1682 | break; | |
1683 | } | |
1684 | ||
1685 | if (index >= htabslots) { | |
1686 | assert(index == htabslots); | |
1687 | index = 0; | |
1688 | } | |
1689 | } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final)); | |
1690 | ||
1691 | if (index >= htabslots) { | |
1692 | assert(index == htabslots); | |
1693 | index = 0; | |
1694 | } | |
1695 | ||
1696 | spapr->htab_save_index = index; | |
1697 | ||
e68cb8b4 | 1698 | return (examined >= htabslots) && (sent == 0) ? 1 : 0; |
4be21d56 DG |
1699 | } |
1700 | ||
e68cb8b4 AK |
1701 | #define MAX_ITERATION_NS 5000000 /* 5 ms */ |
1702 | #define MAX_KVM_BUF_SIZE 2048 | |
1703 | ||
4be21d56 DG |
1704 | static int htab_save_iterate(QEMUFile *f, void *opaque) |
1705 | { | |
28e02042 | 1706 | sPAPRMachineState *spapr = opaque; |
715c5407 | 1707 | int fd; |
e68cb8b4 | 1708 | int rc = 0; |
4be21d56 DG |
1709 | |
1710 | /* Iteration header */ | |
1711 | qemu_put_be32(f, 0); | |
1712 | ||
e68cb8b4 AK |
1713 | if (!spapr->htab) { |
1714 | assert(kvm_enabled()); | |
1715 | ||
715c5407 DG |
1716 | fd = get_htab_fd(spapr); |
1717 | if (fd < 0) { | |
1718 | return fd; | |
01a57972 SMJ |
1719 | } |
1720 | ||
715c5407 | 1721 | rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS); |
e68cb8b4 AK |
1722 | if (rc < 0) { |
1723 | return rc; | |
1724 | } | |
1725 | } else if (spapr->htab_first_pass) { | |
4be21d56 DG |
1726 | htab_save_first_pass(f, spapr, MAX_ITERATION_NS); |
1727 | } else { | |
e68cb8b4 | 1728 | rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS); |
4be21d56 DG |
1729 | } |
1730 | ||
1731 | /* End marker */ | |
1732 | qemu_put_be32(f, 0); | |
1733 | qemu_put_be16(f, 0); | |
1734 | qemu_put_be16(f, 0); | |
1735 | ||
e68cb8b4 | 1736 | return rc; |
4be21d56 DG |
1737 | } |
1738 | ||
1739 | static int htab_save_complete(QEMUFile *f, void *opaque) | |
1740 | { | |
28e02042 | 1741 | sPAPRMachineState *spapr = opaque; |
715c5407 | 1742 | int fd; |
4be21d56 DG |
1743 | |
1744 | /* Iteration header */ | |
1745 | qemu_put_be32(f, 0); | |
1746 | ||
e68cb8b4 AK |
1747 | if (!spapr->htab) { |
1748 | int rc; | |
1749 | ||
1750 | assert(kvm_enabled()); | |
1751 | ||
715c5407 DG |
1752 | fd = get_htab_fd(spapr); |
1753 | if (fd < 0) { | |
1754 | return fd; | |
01a57972 SMJ |
1755 | } |
1756 | ||
715c5407 | 1757 | rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1); |
e68cb8b4 AK |
1758 | if (rc < 0) { |
1759 | return rc; | |
1760 | } | |
e68cb8b4 | 1761 | } else { |
378bc217 DG |
1762 | if (spapr->htab_first_pass) { |
1763 | htab_save_first_pass(f, spapr, -1); | |
1764 | } | |
e68cb8b4 AK |
1765 | htab_save_later_pass(f, spapr, -1); |
1766 | } | |
4be21d56 DG |
1767 | |
1768 | /* End marker */ | |
1769 | qemu_put_be32(f, 0); | |
1770 | qemu_put_be16(f, 0); | |
1771 | qemu_put_be16(f, 0); | |
1772 | ||
1773 | return 0; | |
1774 | } | |
1775 | ||
1776 | static int htab_load(QEMUFile *f, void *opaque, int version_id) | |
1777 | { | |
28e02042 | 1778 | sPAPRMachineState *spapr = opaque; |
4be21d56 | 1779 | uint32_t section_hdr; |
e68cb8b4 | 1780 | int fd = -1; |
4be21d56 DG |
1781 | |
1782 | if (version_id < 1 || version_id > 1) { | |
98a5d100 | 1783 | error_report("htab_load() bad version"); |
4be21d56 DG |
1784 | return -EINVAL; |
1785 | } | |
1786 | ||
1787 | section_hdr = qemu_get_be32(f); | |
1788 | ||
1789 | if (section_hdr) { | |
9897e462 | 1790 | Error *local_err = NULL; |
c5f54f3e DG |
1791 | |
1792 | /* First section gives the htab size */ | |
1793 | spapr_reallocate_hpt(spapr, section_hdr, &local_err); | |
1794 | if (local_err) { | |
1795 | error_report_err(local_err); | |
4be21d56 DG |
1796 | return -EINVAL; |
1797 | } | |
1798 | return 0; | |
1799 | } | |
1800 | ||
e68cb8b4 AK |
1801 | if (!spapr->htab) { |
1802 | assert(kvm_enabled()); | |
1803 | ||
1804 | fd = kvmppc_get_htab_fd(true); | |
1805 | if (fd < 0) { | |
98a5d100 DG |
1806 | error_report("Unable to open fd to restore KVM hash table: %s", |
1807 | strerror(errno)); | |
e68cb8b4 AK |
1808 | } |
1809 | } | |
1810 | ||
4be21d56 DG |
1811 | while (true) { |
1812 | uint32_t index; | |
1813 | uint16_t n_valid, n_invalid; | |
1814 | ||
1815 | index = qemu_get_be32(f); | |
1816 | n_valid = qemu_get_be16(f); | |
1817 | n_invalid = qemu_get_be16(f); | |
1818 | ||
1819 | if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) { | |
1820 | /* End of Stream */ | |
1821 | break; | |
1822 | } | |
1823 | ||
e68cb8b4 | 1824 | if ((index + n_valid + n_invalid) > |
4be21d56 DG |
1825 | (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) { |
1826 | /* Bad index in stream */ | |
98a5d100 DG |
1827 | error_report( |
1828 | "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)", | |
1829 | index, n_valid, n_invalid, spapr->htab_shift); | |
4be21d56 DG |
1830 | return -EINVAL; |
1831 | } | |
1832 | ||
e68cb8b4 AK |
1833 | if (spapr->htab) { |
1834 | if (n_valid) { | |
1835 | qemu_get_buffer(f, HPTE(spapr->htab, index), | |
1836 | HASH_PTE_SIZE_64 * n_valid); | |
1837 | } | |
1838 | if (n_invalid) { | |
1839 | memset(HPTE(spapr->htab, index + n_valid), 0, | |
1840 | HASH_PTE_SIZE_64 * n_invalid); | |
1841 | } | |
1842 | } else { | |
1843 | int rc; | |
1844 | ||
1845 | assert(fd >= 0); | |
1846 | ||
1847 | rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid); | |
1848 | if (rc < 0) { | |
1849 | return rc; | |
1850 | } | |
4be21d56 DG |
1851 | } |
1852 | } | |
1853 | ||
e68cb8b4 AK |
1854 | if (!spapr->htab) { |
1855 | assert(fd >= 0); | |
1856 | close(fd); | |
1857 | } | |
1858 | ||
4be21d56 DG |
1859 | return 0; |
1860 | } | |
1861 | ||
c573fc03 TH |
1862 | static void htab_cleanup(void *opaque) |
1863 | { | |
1864 | sPAPRMachineState *spapr = opaque; | |
1865 | ||
1866 | close_htab_fd(spapr); | |
1867 | } | |
1868 | ||
4be21d56 DG |
1869 | static SaveVMHandlers savevm_htab_handlers = { |
1870 | .save_live_setup = htab_save_setup, | |
1871 | .save_live_iterate = htab_save_iterate, | |
a3e06c3d | 1872 | .save_live_complete_precopy = htab_save_complete, |
c573fc03 | 1873 | .cleanup = htab_cleanup, |
4be21d56 DG |
1874 | .load_state = htab_load, |
1875 | }; | |
1876 | ||
5b2128d2 AG |
1877 | static void spapr_boot_set(void *opaque, const char *boot_device, |
1878 | Error **errp) | |
1879 | { | |
1880 | MachineState *machine = MACHINE(qdev_get_machine()); | |
1881 | machine->boot_order = g_strdup(boot_device); | |
1882 | } | |
1883 | ||
224245bf DG |
1884 | /* |
1885 | * Reset routine for LMB DR devices. | |
1886 | * | |
1887 | * Unlike PCI DR devices, LMB DR devices explicitly register this reset | |
1888 | * routine. Reset for PCI DR devices will be handled by PHB reset routine | |
1889 | * when it walks all its children devices. LMB devices reset occurs | |
1890 | * as part of spapr_ppc_reset(). | |
1891 | */ | |
1892 | static void spapr_drc_reset(void *opaque) | |
1893 | { | |
1894 | sPAPRDRConnector *drc = opaque; | |
1895 | DeviceState *d = DEVICE(drc); | |
1896 | ||
1897 | if (d) { | |
1898 | device_reset(d); | |
1899 | } | |
1900 | } | |
1901 | ||
1902 | static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr) | |
1903 | { | |
1904 | MachineState *machine = MACHINE(spapr); | |
1905 | uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; | |
e8f986fc | 1906 | uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size; |
224245bf DG |
1907 | int i; |
1908 | ||
1909 | for (i = 0; i < nr_lmbs; i++) { | |
1910 | sPAPRDRConnector *drc; | |
1911 | uint64_t addr; | |
1912 | ||
e8f986fc | 1913 | addr = i * lmb_size + spapr->hotplug_memory.base; |
2d335818 | 1914 | drc = spapr_dr_connector_new(OBJECT(spapr), TYPE_SPAPR_DRC_LMB, |
224245bf DG |
1915 | addr/lmb_size); |
1916 | qemu_register_reset(spapr_drc_reset, drc); | |
1917 | } | |
1918 | } | |
1919 | ||
1920 | /* | |
1921 | * If RAM size, maxmem size and individual node mem sizes aren't aligned | |
1922 | * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest | |
1923 | * since we can't support such unaligned sizes with DRCONF_MEMORY. | |
1924 | */ | |
7c150d6f | 1925 | static void spapr_validate_node_memory(MachineState *machine, Error **errp) |
224245bf DG |
1926 | { |
1927 | int i; | |
1928 | ||
7c150d6f DG |
1929 | if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) { |
1930 | error_setg(errp, "Memory size 0x" RAM_ADDR_FMT | |
1931 | " is not aligned to %llu MiB", | |
1932 | machine->ram_size, | |
1933 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
1934 | return; | |
1935 | } | |
1936 | ||
1937 | if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) { | |
1938 | error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT | |
1939 | " is not aligned to %llu MiB", | |
1940 | machine->ram_size, | |
1941 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
1942 | return; | |
224245bf DG |
1943 | } |
1944 | ||
1945 | for (i = 0; i < nb_numa_nodes; i++) { | |
1946 | if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) { | |
7c150d6f DG |
1947 | error_setg(errp, |
1948 | "Node %d memory size 0x%" PRIx64 | |
1949 | " is not aligned to %llu MiB", | |
1950 | i, numa_info[i].node_mem, | |
1951 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
1952 | return; | |
224245bf DG |
1953 | } |
1954 | } | |
1955 | } | |
1956 | ||
535455fd IM |
1957 | /* find cpu slot in machine->possible_cpus by core_id */ |
1958 | static CPUArchId *spapr_find_cpu_slot(MachineState *ms, uint32_t id, int *idx) | |
1959 | { | |
1960 | int index = id / smp_threads; | |
1961 | ||
1962 | if (index >= ms->possible_cpus->len) { | |
1963 | return NULL; | |
1964 | } | |
1965 | if (idx) { | |
1966 | *idx = index; | |
1967 | } | |
1968 | return &ms->possible_cpus->cpus[index]; | |
1969 | } | |
1970 | ||
0c86d0fd DG |
1971 | static void spapr_init_cpus(sPAPRMachineState *spapr) |
1972 | { | |
1973 | MachineState *machine = MACHINE(spapr); | |
1974 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
1975 | char *type = spapr_get_cpu_core_type(machine->cpu_model); | |
1976 | int smt = kvmppc_smt_threads(); | |
535455fd IM |
1977 | const CPUArchIdList *possible_cpus; |
1978 | int boot_cores_nr = smp_cpus / smp_threads; | |
0c86d0fd DG |
1979 | int i; |
1980 | ||
1981 | if (!type) { | |
1982 | error_report("Unable to find sPAPR CPU Core definition"); | |
1983 | exit(1); | |
1984 | } | |
1985 | ||
535455fd | 1986 | possible_cpus = mc->possible_cpu_arch_ids(machine); |
c5514d0e | 1987 | if (mc->has_hotpluggable_cpus) { |
0c86d0fd DG |
1988 | if (smp_cpus % smp_threads) { |
1989 | error_report("smp_cpus (%u) must be multiple of threads (%u)", | |
1990 | smp_cpus, smp_threads); | |
1991 | exit(1); | |
1992 | } | |
1993 | if (max_cpus % smp_threads) { | |
1994 | error_report("max_cpus (%u) must be multiple of threads (%u)", | |
1995 | max_cpus, smp_threads); | |
1996 | exit(1); | |
1997 | } | |
0c86d0fd DG |
1998 | } else { |
1999 | if (max_cpus != smp_cpus) { | |
2000 | error_report("This machine version does not support CPU hotplug"); | |
2001 | exit(1); | |
2002 | } | |
535455fd | 2003 | boot_cores_nr = possible_cpus->len; |
0c86d0fd DG |
2004 | } |
2005 | ||
535455fd | 2006 | for (i = 0; i < possible_cpus->len; i++) { |
0c86d0fd DG |
2007 | int core_id = i * smp_threads; |
2008 | ||
c5514d0e | 2009 | if (mc->has_hotpluggable_cpus) { |
0c86d0fd | 2010 | sPAPRDRConnector *drc = |
2d335818 | 2011 | spapr_dr_connector_new(OBJECT(spapr), TYPE_SPAPR_DRC_CPU, |
0c86d0fd DG |
2012 | (core_id / smp_threads) * smt); |
2013 | ||
2014 | qemu_register_reset(spapr_drc_reset, drc); | |
2015 | } | |
2016 | ||
535455fd | 2017 | if (i < boot_cores_nr) { |
0c86d0fd DG |
2018 | Object *core = object_new(type); |
2019 | int nr_threads = smp_threads; | |
2020 | ||
2021 | /* Handle the partially filled core for older machine types */ | |
2022 | if ((i + 1) * smp_threads >= smp_cpus) { | |
2023 | nr_threads = smp_cpus - i * smp_threads; | |
2024 | } | |
2025 | ||
2026 | object_property_set_int(core, nr_threads, "nr-threads", | |
2027 | &error_fatal); | |
2028 | object_property_set_int(core, core_id, CPU_CORE_PROP_CORE_ID, | |
2029 | &error_fatal); | |
2030 | object_property_set_bool(core, true, "realized", &error_fatal); | |
2031 | } | |
2032 | } | |
2033 | g_free(type); | |
2034 | } | |
2035 | ||
9fdf0c29 | 2036 | /* pSeries LPAR / sPAPR hardware init */ |
3ef96221 | 2037 | static void ppc_spapr_init(MachineState *machine) |
9fdf0c29 | 2038 | { |
28e02042 | 2039 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
224245bf | 2040 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); |
3ef96221 | 2041 | const char *kernel_filename = machine->kernel_filename; |
3ef96221 | 2042 | const char *initrd_filename = machine->initrd_filename; |
8c9f64df | 2043 | PCIHostState *phb; |
9fdf0c29 | 2044 | int i; |
890c2b77 AK |
2045 | MemoryRegion *sysmem = get_system_memory(); |
2046 | MemoryRegion *ram = g_new(MemoryRegion, 1); | |
658fa66b AK |
2047 | MemoryRegion *rma_region; |
2048 | void *rma = NULL; | |
a8170e5e | 2049 | hwaddr rma_alloc_size; |
b082d65a | 2050 | hwaddr node0_size = spapr_node0_size(); |
b7d1f77a | 2051 | long load_limit, fw_size; |
39ac8455 | 2052 | char *filename; |
9fdf0c29 | 2053 | |
226419d6 | 2054 | msi_nonbroken = true; |
0ee2c058 | 2055 | |
d43b45e2 | 2056 | QLIST_INIT(&spapr->phbs); |
0cffce56 | 2057 | QTAILQ_INIT(&spapr->pending_dimm_unplugs); |
d43b45e2 | 2058 | |
354ac20a | 2059 | /* Allocate RMA if necessary */ |
658fa66b | 2060 | rma_alloc_size = kvmppc_alloc_rma(&rma); |
354ac20a DG |
2061 | |
2062 | if (rma_alloc_size == -1) { | |
730fce59 | 2063 | error_report("Unable to create RMA"); |
354ac20a DG |
2064 | exit(1); |
2065 | } | |
7f763a5d | 2066 | |
c4177479 | 2067 | if (rma_alloc_size && (rma_alloc_size < node0_size)) { |
7f763a5d | 2068 | spapr->rma_size = rma_alloc_size; |
354ac20a | 2069 | } else { |
c4177479 | 2070 | spapr->rma_size = node0_size; |
7f763a5d DG |
2071 | |
2072 | /* With KVM, we don't actually know whether KVM supports an | |
2073 | * unbounded RMA (PR KVM) or is limited by the hash table size | |
2074 | * (HV KVM using VRMA), so we always assume the latter | |
2075 | * | |
2076 | * In that case, we also limit the initial allocations for RTAS | |
2077 | * etc... to 256M since we have no way to know what the VRMA size | |
2078 | * is going to be as it depends on the size of the hash table | |
2079 | * isn't determined yet. | |
2080 | */ | |
2081 | if (kvm_enabled()) { | |
2082 | spapr->vrma_adjust = 1; | |
2083 | spapr->rma_size = MIN(spapr->rma_size, 0x10000000); | |
2084 | } | |
912acdf4 BH |
2085 | |
2086 | /* Actually we don't support unbounded RMA anymore since we | |
2087 | * added proper emulation of HV mode. The max we can get is | |
2088 | * 16G which also happens to be what we configure for PAPR | |
2089 | * mode so make sure we don't do anything bigger than that | |
2090 | */ | |
2091 | spapr->rma_size = MIN(spapr->rma_size, 0x400000000ull); | |
354ac20a DG |
2092 | } |
2093 | ||
c4177479 | 2094 | if (spapr->rma_size > node0_size) { |
d54e4d76 DG |
2095 | error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")", |
2096 | spapr->rma_size); | |
c4177479 AK |
2097 | exit(1); |
2098 | } | |
2099 | ||
b7d1f77a BH |
2100 | /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */ |
2101 | load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD; | |
9fdf0c29 | 2102 | |
7b565160 | 2103 | /* Set up Interrupt Controller before we create the VCPUs */ |
71cd4dac | 2104 | xics_system_init(machine, XICS_IRQS_SPAPR, &error_fatal); |
7b565160 | 2105 | |
facdb8b6 MR |
2106 | /* Set up containers for ibm,client-set-architecture negotiated options */ |
2107 | spapr->ov5 = spapr_ovec_new(); | |
2108 | spapr->ov5_cas = spapr_ovec_new(); | |
2109 | ||
224245bf | 2110 | if (smc->dr_lmb_enabled) { |
facdb8b6 | 2111 | spapr_ovec_set(spapr->ov5, OV5_DRCONF_MEMORY); |
7c150d6f | 2112 | spapr_validate_node_memory(machine, &error_fatal); |
224245bf DG |
2113 | } |
2114 | ||
417ece33 | 2115 | spapr_ovec_set(spapr->ov5, OV5_FORM1_AFFINITY); |
545d6e2b SJS |
2116 | if (!kvm_enabled() || kvmppc_has_cap_mmu_radix()) { |
2117 | /* KVM and TCG always allow GTSE with radix... */ | |
9fb4541f SB |
2118 | spapr_ovec_set(spapr->ov5, OV5_MMU_RADIX_GTSE); |
2119 | } | |
2120 | /* ... but not with hash (currently). */ | |
417ece33 | 2121 | |
ffbb1705 MR |
2122 | /* advertise support for dedicated HP event source to guests */ |
2123 | if (spapr->use_hotplug_event_source) { | |
2124 | spapr_ovec_set(spapr->ov5, OV5_HP_EVT); | |
2125 | } | |
2126 | ||
9fdf0c29 | 2127 | /* init CPUs */ |
19fb2c36 | 2128 | if (machine->cpu_model == NULL) { |
3daa4a9f | 2129 | machine->cpu_model = kvm_enabled() ? "host" : smc->tcg_default_cpu; |
9fdf0c29 | 2130 | } |
94a94e4c | 2131 | |
e703d2f7 GK |
2132 | ppc_cpu_parse_features(machine->cpu_model); |
2133 | ||
0c86d0fd | 2134 | spapr_init_cpus(spapr); |
9fdf0c29 | 2135 | |
026bfd89 DG |
2136 | if (kvm_enabled()) { |
2137 | /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */ | |
2138 | kvmppc_enable_logical_ci_hcalls(); | |
ef9971dd | 2139 | kvmppc_enable_set_mode_hcall(); |
5145ad4f NW |
2140 | |
2141 | /* H_CLEAR_MOD/_REF are mandatory in PAPR, but off by default */ | |
2142 | kvmppc_enable_clear_ref_mod_hcalls(); | |
026bfd89 DG |
2143 | } |
2144 | ||
9fdf0c29 | 2145 | /* allocate RAM */ |
f92f5da1 | 2146 | memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram", |
fb164994 | 2147 | machine->ram_size); |
f92f5da1 | 2148 | memory_region_add_subregion(sysmem, 0, ram); |
9fdf0c29 | 2149 | |
658fa66b AK |
2150 | if (rma_alloc_size && rma) { |
2151 | rma_region = g_new(MemoryRegion, 1); | |
2152 | memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma", | |
2153 | rma_alloc_size, rma); | |
2154 | vmstate_register_ram_global(rma_region); | |
2155 | memory_region_add_subregion(sysmem, 0, rma_region); | |
2156 | } | |
2157 | ||
4a1c9cf0 BR |
2158 | /* initialize hotplug memory address space */ |
2159 | if (machine->ram_size < machine->maxram_size) { | |
2160 | ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size; | |
71c9a3dd BR |
2161 | /* |
2162 | * Limit the number of hotpluggable memory slots to half the number | |
2163 | * slots that KVM supports, leaving the other half for PCI and other | |
2164 | * devices. However ensure that number of slots doesn't drop below 32. | |
2165 | */ | |
2166 | int max_memslots = kvm_enabled() ? kvm_get_max_memslots() / 2 : | |
2167 | SPAPR_MAX_RAM_SLOTS; | |
4a1c9cf0 | 2168 | |
71c9a3dd BR |
2169 | if (max_memslots < SPAPR_MAX_RAM_SLOTS) { |
2170 | max_memslots = SPAPR_MAX_RAM_SLOTS; | |
2171 | } | |
2172 | if (machine->ram_slots > max_memslots) { | |
d54e4d76 DG |
2173 | error_report("Specified number of memory slots %" |
2174 | PRIu64" exceeds max supported %d", | |
71c9a3dd | 2175 | machine->ram_slots, max_memslots); |
d54e4d76 | 2176 | exit(1); |
4a1c9cf0 BR |
2177 | } |
2178 | ||
2179 | spapr->hotplug_memory.base = ROUND_UP(machine->ram_size, | |
2180 | SPAPR_HOTPLUG_MEM_ALIGN); | |
2181 | memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr), | |
2182 | "hotplug-memory", hotplug_mem_size); | |
2183 | memory_region_add_subregion(sysmem, spapr->hotplug_memory.base, | |
2184 | &spapr->hotplug_memory.mr); | |
2185 | } | |
2186 | ||
224245bf DG |
2187 | if (smc->dr_lmb_enabled) { |
2188 | spapr_create_lmb_dr_connectors(spapr); | |
2189 | } | |
2190 | ||
39ac8455 | 2191 | filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin"); |
4c56440d | 2192 | if (!filename) { |
730fce59 | 2193 | error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin"); |
4c56440d SW |
2194 | exit(1); |
2195 | } | |
b7d1f77a | 2196 | spapr->rtas_size = get_image_size(filename); |
8afc22a2 ZJ |
2197 | if (spapr->rtas_size < 0) { |
2198 | error_report("Could not get size of LPAR rtas '%s'", filename); | |
2199 | exit(1); | |
2200 | } | |
b7d1f77a BH |
2201 | spapr->rtas_blob = g_malloc(spapr->rtas_size); |
2202 | if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) { | |
730fce59 | 2203 | error_report("Could not load LPAR rtas '%s'", filename); |
39ac8455 DG |
2204 | exit(1); |
2205 | } | |
4d8d5467 | 2206 | if (spapr->rtas_size > RTAS_MAX_SIZE) { |
730fce59 TH |
2207 | error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)", |
2208 | (size_t)spapr->rtas_size, RTAS_MAX_SIZE); | |
4d8d5467 BH |
2209 | exit(1); |
2210 | } | |
7267c094 | 2211 | g_free(filename); |
39ac8455 | 2212 | |
ffbb1705 | 2213 | /* Set up RTAS event infrastructure */ |
74d042e5 DG |
2214 | spapr_events_init(spapr); |
2215 | ||
12f42174 | 2216 | /* Set up the RTC RTAS interfaces */ |
28df36a1 | 2217 | spapr_rtc_create(spapr); |
12f42174 | 2218 | |
b5cec4c5 | 2219 | /* Set up VIO bus */ |
4040ab72 DG |
2220 | spapr->vio_bus = spapr_vio_bus_init(); |
2221 | ||
277f9acf | 2222 | for (i = 0; i < MAX_SERIAL_PORTS; i++) { |
4040ab72 | 2223 | if (serial_hds[i]) { |
d601fac4 | 2224 | spapr_vty_create(spapr->vio_bus, serial_hds[i]); |
4040ab72 DG |
2225 | } |
2226 | } | |
9fdf0c29 | 2227 | |
639e8102 DG |
2228 | /* We always have at least the nvram device on VIO */ |
2229 | spapr_create_nvram(spapr); | |
2230 | ||
3384f95c | 2231 | /* Set up PCI */ |
fa28f71b AK |
2232 | spapr_pci_rtas_init(); |
2233 | ||
89dfd6e1 | 2234 | phb = spapr_create_phb(spapr, 0); |
3384f95c | 2235 | |
277f9acf | 2236 | for (i = 0; i < nb_nics; i++) { |
8d90ad90 DG |
2237 | NICInfo *nd = &nd_table[i]; |
2238 | ||
2239 | if (!nd->model) { | |
7267c094 | 2240 | nd->model = g_strdup("ibmveth"); |
8d90ad90 DG |
2241 | } |
2242 | ||
2243 | if (strcmp(nd->model, "ibmveth") == 0) { | |
d601fac4 | 2244 | spapr_vlan_create(spapr->vio_bus, nd); |
8d90ad90 | 2245 | } else { |
29b358f9 | 2246 | pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL); |
8d90ad90 DG |
2247 | } |
2248 | } | |
2249 | ||
6e270446 | 2250 | for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) { |
d601fac4 | 2251 | spapr_vscsi_create(spapr->vio_bus); |
6e270446 BH |
2252 | } |
2253 | ||
f28359d8 | 2254 | /* Graphics */ |
14c6a894 | 2255 | if (spapr_vga_init(phb->bus, &error_fatal)) { |
3fc5acde | 2256 | spapr->has_graphics = true; |
c6e76503 | 2257 | machine->usb |= defaults_enabled() && !machine->usb_disabled; |
f28359d8 LZ |
2258 | } |
2259 | ||
4ee9ced9 | 2260 | if (machine->usb) { |
57040d45 TH |
2261 | if (smc->use_ohci_by_default) { |
2262 | pci_create_simple(phb->bus, -1, "pci-ohci"); | |
2263 | } else { | |
2264 | pci_create_simple(phb->bus, -1, "nec-usb-xhci"); | |
2265 | } | |
c86580b8 | 2266 | |
35139a59 | 2267 | if (spapr->has_graphics) { |
c86580b8 MA |
2268 | USBBus *usb_bus = usb_bus_find(-1); |
2269 | ||
2270 | usb_create_simple(usb_bus, "usb-kbd"); | |
2271 | usb_create_simple(usb_bus, "usb-mouse"); | |
35139a59 DG |
2272 | } |
2273 | } | |
2274 | ||
7f763a5d | 2275 | if (spapr->rma_size < (MIN_RMA_SLOF << 20)) { |
d54e4d76 DG |
2276 | error_report( |
2277 | "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)", | |
2278 | MIN_RMA_SLOF); | |
4d8d5467 BH |
2279 | exit(1); |
2280 | } | |
2281 | ||
9fdf0c29 DG |
2282 | if (kernel_filename) { |
2283 | uint64_t lowaddr = 0; | |
2284 | ||
a19f7fb0 DG |
2285 | spapr->kernel_size = load_elf(kernel_filename, translate_kernel_address, |
2286 | NULL, NULL, &lowaddr, NULL, 1, | |
2287 | PPC_ELF_MACHINE, 0, 0); | |
2288 | if (spapr->kernel_size == ELF_LOAD_WRONG_ENDIAN) { | |
2289 | spapr->kernel_size = load_elf(kernel_filename, | |
2290 | translate_kernel_address, NULL, NULL, | |
2291 | &lowaddr, NULL, 0, PPC_ELF_MACHINE, | |
2292 | 0, 0); | |
2293 | spapr->kernel_le = spapr->kernel_size > 0; | |
16457e7f | 2294 | } |
a19f7fb0 DG |
2295 | if (spapr->kernel_size < 0) { |
2296 | error_report("error loading %s: %s", kernel_filename, | |
2297 | load_elf_strerror(spapr->kernel_size)); | |
9fdf0c29 DG |
2298 | exit(1); |
2299 | } | |
2300 | ||
2301 | /* load initrd */ | |
2302 | if (initrd_filename) { | |
4d8d5467 BH |
2303 | /* Try to locate the initrd in the gap between the kernel |
2304 | * and the firmware. Add a bit of space just in case | |
2305 | */ | |
a19f7fb0 DG |
2306 | spapr->initrd_base = (KERNEL_LOAD_ADDR + spapr->kernel_size |
2307 | + 0x1ffff) & ~0xffff; | |
2308 | spapr->initrd_size = load_image_targphys(initrd_filename, | |
2309 | spapr->initrd_base, | |
2310 | load_limit | |
2311 | - spapr->initrd_base); | |
2312 | if (spapr->initrd_size < 0) { | |
d54e4d76 DG |
2313 | error_report("could not load initial ram disk '%s'", |
2314 | initrd_filename); | |
9fdf0c29 DG |
2315 | exit(1); |
2316 | } | |
9fdf0c29 | 2317 | } |
4d8d5467 | 2318 | } |
a3467baa | 2319 | |
8e7ea787 AF |
2320 | if (bios_name == NULL) { |
2321 | bios_name = FW_FILE_NAME; | |
2322 | } | |
2323 | filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name); | |
4c56440d | 2324 | if (!filename) { |
68fea5a0 | 2325 | error_report("Could not find LPAR firmware '%s'", bios_name); |
4c56440d SW |
2326 | exit(1); |
2327 | } | |
4d8d5467 | 2328 | fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE); |
68fea5a0 TH |
2329 | if (fw_size <= 0) { |
2330 | error_report("Could not load LPAR firmware '%s'", filename); | |
4d8d5467 BH |
2331 | exit(1); |
2332 | } | |
2333 | g_free(filename); | |
4d8d5467 | 2334 | |
28e02042 DG |
2335 | /* FIXME: Should register things through the MachineState's qdev |
2336 | * interface, this is a legacy from the sPAPREnvironment structure | |
2337 | * which predated MachineState but had a similar function */ | |
4be21d56 DG |
2338 | vmstate_register(NULL, 0, &vmstate_spapr, spapr); |
2339 | register_savevm_live(NULL, "spapr/htab", -1, 1, | |
2340 | &savevm_htab_handlers, spapr); | |
2341 | ||
5b2128d2 | 2342 | qemu_register_boot_set(spapr_boot_set, spapr); |
42043e4f | 2343 | |
42043e4f | 2344 | if (kvm_enabled()) { |
3dc410ae | 2345 | /* to stop and start vmclock */ |
42043e4f LV |
2346 | qemu_add_vm_change_state_handler(cpu_ppc_clock_vm_state_change, |
2347 | &spapr->tb); | |
3dc410ae AK |
2348 | |
2349 | kvmppc_spapr_enable_inkernel_multitce(); | |
42043e4f | 2350 | } |
9fdf0c29 DG |
2351 | } |
2352 | ||
135a129a AK |
2353 | static int spapr_kvm_type(const char *vm_type) |
2354 | { | |
2355 | if (!vm_type) { | |
2356 | return 0; | |
2357 | } | |
2358 | ||
2359 | if (!strcmp(vm_type, "HV")) { | |
2360 | return 1; | |
2361 | } | |
2362 | ||
2363 | if (!strcmp(vm_type, "PR")) { | |
2364 | return 2; | |
2365 | } | |
2366 | ||
2367 | error_report("Unknown kvm-type specified '%s'", vm_type); | |
2368 | exit(1); | |
2369 | } | |
2370 | ||
71461b0f | 2371 | /* |
627b84f4 | 2372 | * Implementation of an interface to adjust firmware path |
71461b0f AK |
2373 | * for the bootindex property handling. |
2374 | */ | |
2375 | static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus, | |
2376 | DeviceState *dev) | |
2377 | { | |
2378 | #define CAST(type, obj, name) \ | |
2379 | ((type *)object_dynamic_cast(OBJECT(obj), (name))) | |
2380 | SCSIDevice *d = CAST(SCSIDevice, dev, TYPE_SCSI_DEVICE); | |
2381 | sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE); | |
c4e13492 | 2382 | VHostSCSICommon *vsc = CAST(VHostSCSICommon, dev, TYPE_VHOST_SCSI_COMMON); |
71461b0f AK |
2383 | |
2384 | if (d) { | |
2385 | void *spapr = CAST(void, bus->parent, "spapr-vscsi"); | |
2386 | VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI); | |
2387 | USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE); | |
2388 | ||
2389 | if (spapr) { | |
2390 | /* | |
2391 | * Replace "channel@0/disk@0,0" with "disk@8000000000000000": | |
2392 | * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun | |
2393 | * in the top 16 bits of the 64-bit LUN | |
2394 | */ | |
2395 | unsigned id = 0x8000 | (d->id << 8) | d->lun; | |
2396 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2397 | (uint64_t)id << 48); | |
2398 | } else if (virtio) { | |
2399 | /* | |
2400 | * We use SRP luns of the form 01000000 | (target << 8) | lun | |
2401 | * in the top 32 bits of the 64-bit LUN | |
2402 | * Note: the quote above is from SLOF and it is wrong, | |
2403 | * the actual binding is: | |
2404 | * swap 0100 or 10 << or 20 << ( target lun-id -- srplun ) | |
2405 | */ | |
2406 | unsigned id = 0x1000000 | (d->id << 16) | d->lun; | |
2407 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2408 | (uint64_t)id << 32); | |
2409 | } else if (usb) { | |
2410 | /* | |
2411 | * We use SRP luns of the form 01000000 | (usb-port << 16) | lun | |
2412 | * in the top 32 bits of the 64-bit LUN | |
2413 | */ | |
2414 | unsigned usb_port = atoi(usb->port->path); | |
2415 | unsigned id = 0x1000000 | (usb_port << 16) | d->lun; | |
2416 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2417 | (uint64_t)id << 32); | |
2418 | } | |
2419 | } | |
2420 | ||
b99260eb TH |
2421 | /* |
2422 | * SLOF probes the USB devices, and if it recognizes that the device is a | |
2423 | * storage device, it changes its name to "storage" instead of "usb-host", | |
2424 | * and additionally adds a child node for the SCSI LUN, so the correct | |
2425 | * boot path in SLOF is something like .../storage@1/disk@xxx" instead. | |
2426 | */ | |
2427 | if (strcmp("usb-host", qdev_fw_name(dev)) == 0) { | |
2428 | USBDevice *usbdev = CAST(USBDevice, dev, TYPE_USB_DEVICE); | |
2429 | if (usb_host_dev_is_scsi_storage(usbdev)) { | |
2430 | return g_strdup_printf("storage@%s/disk", usbdev->port->path); | |
2431 | } | |
2432 | } | |
2433 | ||
71461b0f AK |
2434 | if (phb) { |
2435 | /* Replace "pci" with "pci@800000020000000" */ | |
2436 | return g_strdup_printf("pci@%"PRIX64, phb->buid); | |
2437 | } | |
2438 | ||
c4e13492 FF |
2439 | if (vsc) { |
2440 | /* Same logic as virtio above */ | |
2441 | unsigned id = 0x1000000 | (vsc->target << 16) | vsc->lun; | |
2442 | return g_strdup_printf("disk@%"PRIX64, (uint64_t)id << 32); | |
2443 | } | |
2444 | ||
71461b0f AK |
2445 | return NULL; |
2446 | } | |
2447 | ||
23825581 EH |
2448 | static char *spapr_get_kvm_type(Object *obj, Error **errp) |
2449 | { | |
28e02042 | 2450 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
23825581 | 2451 | |
28e02042 | 2452 | return g_strdup(spapr->kvm_type); |
23825581 EH |
2453 | } |
2454 | ||
2455 | static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp) | |
2456 | { | |
28e02042 | 2457 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
23825581 | 2458 | |
28e02042 DG |
2459 | g_free(spapr->kvm_type); |
2460 | spapr->kvm_type = g_strdup(value); | |
23825581 EH |
2461 | } |
2462 | ||
f6229214 MR |
2463 | static bool spapr_get_modern_hotplug_events(Object *obj, Error **errp) |
2464 | { | |
2465 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2466 | ||
2467 | return spapr->use_hotplug_event_source; | |
2468 | } | |
2469 | ||
2470 | static void spapr_set_modern_hotplug_events(Object *obj, bool value, | |
2471 | Error **errp) | |
2472 | { | |
2473 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2474 | ||
2475 | spapr->use_hotplug_event_source = value; | |
2476 | } | |
2477 | ||
23825581 EH |
2478 | static void spapr_machine_initfn(Object *obj) |
2479 | { | |
715c5407 DG |
2480 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
2481 | ||
2482 | spapr->htab_fd = -1; | |
f6229214 | 2483 | spapr->use_hotplug_event_source = true; |
23825581 EH |
2484 | object_property_add_str(obj, "kvm-type", |
2485 | spapr_get_kvm_type, spapr_set_kvm_type, NULL); | |
49d2e648 MA |
2486 | object_property_set_description(obj, "kvm-type", |
2487 | "Specifies the KVM virtualization mode (HV, PR)", | |
2488 | NULL); | |
f6229214 MR |
2489 | object_property_add_bool(obj, "modern-hotplug-events", |
2490 | spapr_get_modern_hotplug_events, | |
2491 | spapr_set_modern_hotplug_events, | |
2492 | NULL); | |
2493 | object_property_set_description(obj, "modern-hotplug-events", | |
2494 | "Use dedicated hotplug event mechanism in" | |
2495 | " place of standard EPOW events when possible" | |
2496 | " (required for memory hot-unplug support)", | |
2497 | NULL); | |
23825581 EH |
2498 | } |
2499 | ||
87bbdd9c DG |
2500 | static void spapr_machine_finalizefn(Object *obj) |
2501 | { | |
2502 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2503 | ||
2504 | g_free(spapr->kvm_type); | |
2505 | } | |
2506 | ||
1c7ad77e | 2507 | void spapr_do_system_reset_on_cpu(CPUState *cs, run_on_cpu_data arg) |
34316482 | 2508 | { |
34316482 AK |
2509 | cpu_synchronize_state(cs); |
2510 | ppc_cpu_do_system_reset(cs); | |
2511 | } | |
2512 | ||
2513 | static void spapr_nmi(NMIState *n, int cpu_index, Error **errp) | |
2514 | { | |
2515 | CPUState *cs; | |
2516 | ||
2517 | CPU_FOREACH(cs) { | |
1c7ad77e | 2518 | async_run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL); |
34316482 AK |
2519 | } |
2520 | } | |
2521 | ||
79b78a6b MR |
2522 | static void spapr_add_lmbs(DeviceState *dev, uint64_t addr_start, uint64_t size, |
2523 | uint32_t node, bool dedicated_hp_event_source, | |
2524 | Error **errp) | |
c20d332a BR |
2525 | { |
2526 | sPAPRDRConnector *drc; | |
2527 | sPAPRDRConnectorClass *drck; | |
2528 | uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE; | |
2529 | int i, fdt_offset, fdt_size; | |
2530 | void *fdt; | |
79b78a6b | 2531 | uint64_t addr = addr_start; |
c20d332a | 2532 | |
c20d332a | 2533 | for (i = 0; i < nr_lmbs; i++) { |
fbf55397 DG |
2534 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2535 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
c20d332a BR |
2536 | g_assert(drc); |
2537 | ||
2538 | fdt = create_device_tree(&fdt_size); | |
2539 | fdt_offset = spapr_populate_memory_node(fdt, node, addr, | |
2540 | SPAPR_MEMORY_BLOCK_SIZE); | |
2541 | ||
2542 | drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); | |
2543 | drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, errp); | |
c20d332a | 2544 | addr += SPAPR_MEMORY_BLOCK_SIZE; |
5c0139a8 MR |
2545 | if (!dev->hotplugged) { |
2546 | /* guests expect coldplugged LMBs to be pre-allocated */ | |
2547 | drck->set_allocation_state(drc, SPAPR_DR_ALLOCATION_STATE_USABLE); | |
2548 | drck->set_isolation_state(drc, SPAPR_DR_ISOLATION_STATE_UNISOLATED); | |
2549 | } | |
c20d332a | 2550 | } |
5dd5238c JD |
2551 | /* send hotplug notification to the |
2552 | * guest only in case of hotplugged memory | |
2553 | */ | |
2554 | if (dev->hotplugged) { | |
79b78a6b | 2555 | if (dedicated_hp_event_source) { |
fbf55397 DG |
2556 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2557 | addr_start / SPAPR_MEMORY_BLOCK_SIZE); | |
79b78a6b MR |
2558 | drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); |
2559 | spapr_hotplug_req_add_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, | |
2560 | nr_lmbs, | |
0b55aa91 | 2561 | spapr_drc_index(drc)); |
79b78a6b MR |
2562 | } else { |
2563 | spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, | |
2564 | nr_lmbs); | |
2565 | } | |
5dd5238c | 2566 | } |
c20d332a BR |
2567 | } |
2568 | ||
2569 | static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
2570 | uint32_t node, Error **errp) | |
2571 | { | |
2572 | Error *local_err = NULL; | |
2573 | sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev); | |
2574 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2575 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
2576 | MemoryRegion *mr = ddc->get_memory_region(dimm); | |
2577 | uint64_t align = memory_region_get_alignment(mr); | |
2578 | uint64_t size = memory_region_size(mr); | |
2579 | uint64_t addr; | |
df587133 | 2580 | |
d6a9b0b8 | 2581 | pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err); |
c20d332a BR |
2582 | if (local_err) { |
2583 | goto out; | |
2584 | } | |
2585 | ||
2586 | addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err); | |
2587 | if (local_err) { | |
2588 | pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr); | |
2589 | goto out; | |
2590 | } | |
2591 | ||
79b78a6b MR |
2592 | spapr_add_lmbs(dev, addr, size, node, |
2593 | spapr_ovec_test(ms->ov5_cas, OV5_HP_EVT), | |
2594 | &error_abort); | |
c20d332a BR |
2595 | |
2596 | out: | |
2597 | error_propagate(errp, local_err); | |
2598 | } | |
2599 | ||
c871bc70 LV |
2600 | static void spapr_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev, |
2601 | Error **errp) | |
2602 | { | |
2603 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2604 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
2605 | MemoryRegion *mr = ddc->get_memory_region(dimm); | |
2606 | uint64_t size = memory_region_size(mr); | |
2607 | char *mem_dev; | |
2608 | ||
2609 | if (size % SPAPR_MEMORY_BLOCK_SIZE) { | |
2610 | error_setg(errp, "Hotplugged memory size must be a multiple of " | |
2611 | "%lld MB", SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
2612 | return; | |
2613 | } | |
2614 | ||
2615 | mem_dev = object_property_get_str(OBJECT(dimm), PC_DIMM_MEMDEV_PROP, NULL); | |
2616 | if (mem_dev && !kvmppc_is_mem_backend_page_size_ok(mem_dev)) { | |
2617 | error_setg(errp, "Memory backend has bad page size. " | |
2618 | "Use 'memory-backend-file' with correct mem-path."); | |
2619 | return; | |
2620 | } | |
2621 | } | |
2622 | ||
0cffce56 DG |
2623 | struct sPAPRDIMMState { |
2624 | PCDIMMDevice *dimm; | |
cf632463 | 2625 | uint32_t nr_lmbs; |
0cffce56 DG |
2626 | QTAILQ_ENTRY(sPAPRDIMMState) next; |
2627 | }; | |
2628 | ||
2629 | static sPAPRDIMMState *spapr_pending_dimm_unplugs_find(sPAPRMachineState *s, | |
2630 | PCDIMMDevice *dimm) | |
2631 | { | |
2632 | sPAPRDIMMState *dimm_state = NULL; | |
2633 | ||
2634 | QTAILQ_FOREACH(dimm_state, &s->pending_dimm_unplugs, next) { | |
2635 | if (dimm_state->dimm == dimm) { | |
2636 | break; | |
2637 | } | |
2638 | } | |
2639 | return dimm_state; | |
2640 | } | |
2641 | ||
2642 | static void spapr_pending_dimm_unplugs_add(sPAPRMachineState *spapr, | |
2643 | sPAPRDIMMState *dimm_state) | |
2644 | { | |
2645 | g_assert(!spapr_pending_dimm_unplugs_find(spapr, dimm_state->dimm)); | |
2646 | QTAILQ_INSERT_HEAD(&spapr->pending_dimm_unplugs, dimm_state, next); | |
2647 | } | |
2648 | ||
2649 | static void spapr_pending_dimm_unplugs_remove(sPAPRMachineState *spapr, | |
2650 | sPAPRDIMMState *dimm_state) | |
2651 | { | |
2652 | QTAILQ_REMOVE(&spapr->pending_dimm_unplugs, dimm_state, next); | |
2653 | g_free(dimm_state); | |
2654 | } | |
cf632463 | 2655 | |
16ee9980 DHB |
2656 | static sPAPRDIMMState *spapr_recover_pending_dimm_state(sPAPRMachineState *ms, |
2657 | PCDIMMDevice *dimm) | |
2658 | { | |
2659 | sPAPRDRConnector *drc; | |
2660 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
2661 | MemoryRegion *mr = ddc->get_memory_region(dimm); | |
2662 | uint64_t size = memory_region_size(mr); | |
2663 | uint32_t nr_lmbs = size / SPAPR_MEMORY_BLOCK_SIZE; | |
2664 | uint32_t avail_lmbs = 0; | |
2665 | uint64_t addr_start, addr; | |
2666 | int i; | |
2667 | sPAPRDIMMState *ds; | |
2668 | ||
2669 | addr_start = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, | |
2670 | &error_abort); | |
2671 | ||
2672 | addr = addr_start; | |
2673 | for (i = 0; i < nr_lmbs; i++) { | |
fbf55397 DG |
2674 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2675 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
16ee9980 DHB |
2676 | g_assert(drc); |
2677 | if (drc->indicator_state != SPAPR_DR_INDICATOR_STATE_INACTIVE) { | |
2678 | avail_lmbs++; | |
2679 | } | |
2680 | addr += SPAPR_MEMORY_BLOCK_SIZE; | |
2681 | } | |
2682 | ||
2683 | ds = g_malloc0(sizeof(sPAPRDIMMState)); | |
2684 | ds->nr_lmbs = avail_lmbs; | |
2685 | ds->dimm = dimm; | |
2686 | spapr_pending_dimm_unplugs_add(ms, ds); | |
2687 | return ds; | |
2688 | } | |
2689 | ||
31834723 DHB |
2690 | /* Callback to be called during DRC release. */ |
2691 | void spapr_lmb_release(DeviceState *dev) | |
cf632463 | 2692 | { |
0cffce56 DG |
2693 | HotplugHandler *hotplug_ctrl = qdev_get_hotplug_handler(dev); |
2694 | sPAPRMachineState *spapr = SPAPR_MACHINE(hotplug_ctrl); | |
2695 | sPAPRDIMMState *ds = spapr_pending_dimm_unplugs_find(spapr, PC_DIMM(dev)); | |
cf632463 | 2696 | |
16ee9980 DHB |
2697 | /* This information will get lost if a migration occurs |
2698 | * during the unplug process. In this case recover it. */ | |
2699 | if (ds == NULL) { | |
2700 | ds = spapr_recover_pending_dimm_state(spapr, PC_DIMM(dev)); | |
2701 | if (ds->nr_lmbs) { | |
2702 | return; | |
2703 | } | |
2704 | } else if (--ds->nr_lmbs) { | |
cf632463 BR |
2705 | return; |
2706 | } | |
2707 | ||
0cffce56 | 2708 | spapr_pending_dimm_unplugs_remove(spapr, ds); |
cf632463 BR |
2709 | |
2710 | /* | |
2711 | * Now that all the LMBs have been removed by the guest, call the | |
2712 | * pc-dimm unplug handler to cleanup up the pc-dimm device. | |
2713 | */ | |
cf632463 BR |
2714 | hotplug_handler_unplug(hotplug_ctrl, dev, &error_abort); |
2715 | } | |
2716 | ||
cf632463 BR |
2717 | static void spapr_memory_unplug(HotplugHandler *hotplug_dev, DeviceState *dev, |
2718 | Error **errp) | |
2719 | { | |
2720 | sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev); | |
2721 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2722 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
2723 | MemoryRegion *mr = ddc->get_memory_region(dimm); | |
2724 | ||
2725 | pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr); | |
2726 | object_unparent(OBJECT(dev)); | |
2727 | } | |
2728 | ||
2729 | static void spapr_memory_unplug_request(HotplugHandler *hotplug_dev, | |
2730 | DeviceState *dev, Error **errp) | |
2731 | { | |
0cffce56 | 2732 | sPAPRMachineState *spapr = SPAPR_MACHINE(hotplug_dev); |
cf632463 BR |
2733 | Error *local_err = NULL; |
2734 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2735 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
2736 | MemoryRegion *mr = ddc->get_memory_region(dimm); | |
2737 | uint64_t size = memory_region_size(mr); | |
0cffce56 DG |
2738 | uint32_t nr_lmbs = size / SPAPR_MEMORY_BLOCK_SIZE; |
2739 | uint64_t addr_start, addr; | |
2740 | int i; | |
2741 | sPAPRDRConnector *drc; | |
2742 | sPAPRDRConnectorClass *drck; | |
2743 | sPAPRDIMMState *ds; | |
cf632463 | 2744 | |
0cffce56 DG |
2745 | addr_start = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, |
2746 | &local_err); | |
cf632463 BR |
2747 | if (local_err) { |
2748 | goto out; | |
2749 | } | |
2750 | ||
0cffce56 DG |
2751 | ds = g_malloc0(sizeof(sPAPRDIMMState)); |
2752 | ds->nr_lmbs = nr_lmbs; | |
2753 | ds->dimm = dimm; | |
2754 | spapr_pending_dimm_unplugs_add(spapr, ds); | |
2755 | ||
2756 | addr = addr_start; | |
2757 | for (i = 0; i < nr_lmbs; i++) { | |
fbf55397 DG |
2758 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2759 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
0cffce56 DG |
2760 | g_assert(drc); |
2761 | ||
2762 | drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); | |
31834723 | 2763 | drck->detach(drc, dev, errp); |
0cffce56 DG |
2764 | addr += SPAPR_MEMORY_BLOCK_SIZE; |
2765 | } | |
2766 | ||
fbf55397 DG |
2767 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2768 | addr_start / SPAPR_MEMORY_BLOCK_SIZE); | |
0cffce56 DG |
2769 | drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); |
2770 | spapr_hotplug_req_remove_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, | |
0b55aa91 | 2771 | nr_lmbs, spapr_drc_index(drc)); |
cf632463 BR |
2772 | out: |
2773 | error_propagate(errp, local_err); | |
2774 | } | |
2775 | ||
af81cf32 BR |
2776 | void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset, |
2777 | sPAPRMachineState *spapr) | |
2778 | { | |
2779 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
2780 | DeviceClass *dc = DEVICE_GET_CLASS(cs); | |
2781 | int id = ppc_get_vcpu_dt_id(cpu); | |
2782 | void *fdt; | |
2783 | int offset, fdt_size; | |
2784 | char *nodename; | |
2785 | ||
2786 | fdt = create_device_tree(&fdt_size); | |
2787 | nodename = g_strdup_printf("%s@%x", dc->fw_name, id); | |
2788 | offset = fdt_add_subnode(fdt, 0, nodename); | |
2789 | ||
2790 | spapr_populate_cpu_dt(cs, fdt, offset, spapr); | |
2791 | g_free(nodename); | |
2792 | ||
2793 | *fdt_offset = offset; | |
2794 | return fdt; | |
2795 | } | |
2796 | ||
115debf2 IM |
2797 | static void spapr_core_unplug(HotplugHandler *hotplug_dev, DeviceState *dev, |
2798 | Error **errp) | |
ff9006dd | 2799 | { |
535455fd | 2800 | MachineState *ms = MACHINE(qdev_get_machine()); |
ff9006dd | 2801 | CPUCore *cc = CPU_CORE(dev); |
535455fd | 2802 | CPUArchId *core_slot = spapr_find_cpu_slot(ms, cc->core_id, NULL); |
ff9006dd | 2803 | |
07572c06 | 2804 | assert(core_slot); |
535455fd | 2805 | core_slot->cpu = NULL; |
ff9006dd IM |
2806 | object_unparent(OBJECT(dev)); |
2807 | } | |
2808 | ||
31834723 DHB |
2809 | /* Callback to be called during DRC release. */ |
2810 | void spapr_core_release(DeviceState *dev) | |
115debf2 IM |
2811 | { |
2812 | HotplugHandler *hotplug_ctrl; | |
2813 | ||
2814 | hotplug_ctrl = qdev_get_hotplug_handler(dev); | |
2815 | hotplug_handler_unplug(hotplug_ctrl, dev, &error_abort); | |
2816 | } | |
2817 | ||
2818 | static | |
2819 | void spapr_core_unplug_request(HotplugHandler *hotplug_dev, DeviceState *dev, | |
2820 | Error **errp) | |
ff9006dd | 2821 | { |
535455fd IM |
2822 | int index; |
2823 | sPAPRDRConnector *drc; | |
ff9006dd IM |
2824 | sPAPRDRConnectorClass *drck; |
2825 | Error *local_err = NULL; | |
535455fd IM |
2826 | CPUCore *cc = CPU_CORE(dev); |
2827 | int smt = kvmppc_smt_threads(); | |
ff9006dd | 2828 | |
535455fd IM |
2829 | if (!spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index)) { |
2830 | error_setg(errp, "Unable to find CPU core with core-id: %d", | |
2831 | cc->core_id); | |
2832 | return; | |
2833 | } | |
ff9006dd IM |
2834 | if (index == 0) { |
2835 | error_setg(errp, "Boot CPU core may not be unplugged"); | |
2836 | return; | |
2837 | } | |
2838 | ||
fbf55397 | 2839 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index * smt); |
ff9006dd IM |
2840 | g_assert(drc); |
2841 | ||
2842 | drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); | |
31834723 | 2843 | drck->detach(drc, dev, &local_err); |
ff9006dd IM |
2844 | if (local_err) { |
2845 | error_propagate(errp, local_err); | |
2846 | return; | |
2847 | } | |
2848 | ||
2849 | spapr_hotplug_req_remove_by_index(drc); | |
2850 | } | |
2851 | ||
2852 | static void spapr_core_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
2853 | Error **errp) | |
2854 | { | |
2855 | sPAPRMachineState *spapr = SPAPR_MACHINE(OBJECT(hotplug_dev)); | |
2856 | MachineClass *mc = MACHINE_GET_CLASS(spapr); | |
2857 | sPAPRCPUCore *core = SPAPR_CPU_CORE(OBJECT(dev)); | |
2858 | CPUCore *cc = CPU_CORE(dev); | |
2859 | CPUState *cs = CPU(core->threads); | |
2860 | sPAPRDRConnector *drc; | |
2861 | Error *local_err = NULL; | |
2862 | void *fdt = NULL; | |
2863 | int fdt_offset = 0; | |
ff9006dd | 2864 | int smt = kvmppc_smt_threads(); |
535455fd IM |
2865 | CPUArchId *core_slot; |
2866 | int index; | |
ff9006dd | 2867 | |
535455fd IM |
2868 | core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); |
2869 | if (!core_slot) { | |
2870 | error_setg(errp, "Unable to find CPU core with core-id: %d", | |
2871 | cc->core_id); | |
2872 | return; | |
2873 | } | |
fbf55397 | 2874 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index * smt); |
ff9006dd | 2875 | |
c5514d0e | 2876 | g_assert(drc || !mc->has_hotpluggable_cpus); |
ff9006dd IM |
2877 | |
2878 | /* | |
2879 | * Setup CPU DT entries only for hotplugged CPUs. For boot time or | |
2880 | * coldplugged CPUs DT entries are setup in spapr_build_fdt(). | |
2881 | */ | |
2882 | if (dev->hotplugged) { | |
2883 | fdt = spapr_populate_hotplug_cpu_dt(cs, &fdt_offset, spapr); | |
2884 | } | |
2885 | ||
2886 | if (drc) { | |
2887 | sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); | |
2888 | drck->attach(drc, dev, fdt, fdt_offset, !dev->hotplugged, &local_err); | |
2889 | if (local_err) { | |
2890 | g_free(fdt); | |
ff9006dd IM |
2891 | error_propagate(errp, local_err); |
2892 | return; | |
2893 | } | |
2894 | } | |
2895 | ||
2896 | if (dev->hotplugged) { | |
2897 | /* | |
2898 | * Send hotplug notification interrupt to the guest only in case | |
2899 | * of hotplugged CPUs. | |
2900 | */ | |
2901 | spapr_hotplug_req_add_by_index(drc); | |
2902 | } else { | |
2903 | /* | |
2904 | * Set the right DRC states for cold plugged CPU. | |
2905 | */ | |
2906 | if (drc) { | |
2907 | sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc); | |
2908 | drck->set_allocation_state(drc, SPAPR_DR_ALLOCATION_STATE_USABLE); | |
2909 | drck->set_isolation_state(drc, SPAPR_DR_ISOLATION_STATE_UNISOLATED); | |
2910 | } | |
2911 | } | |
535455fd | 2912 | core_slot->cpu = OBJECT(dev); |
ff9006dd IM |
2913 | } |
2914 | ||
2915 | static void spapr_core_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
2916 | Error **errp) | |
2917 | { | |
2918 | MachineState *machine = MACHINE(OBJECT(hotplug_dev)); | |
2919 | MachineClass *mc = MACHINE_GET_CLASS(hotplug_dev); | |
ff9006dd IM |
2920 | Error *local_err = NULL; |
2921 | CPUCore *cc = CPU_CORE(dev); | |
2922 | char *base_core_type = spapr_get_cpu_core_type(machine->cpu_model); | |
2923 | const char *type = object_get_typename(OBJECT(dev)); | |
535455fd IM |
2924 | CPUArchId *core_slot; |
2925 | int index; | |
ff9006dd | 2926 | |
c5514d0e | 2927 | if (dev->hotplugged && !mc->has_hotpluggable_cpus) { |
ff9006dd IM |
2928 | error_setg(&local_err, "CPU hotplug not supported for this machine"); |
2929 | goto out; | |
2930 | } | |
2931 | ||
2932 | if (strcmp(base_core_type, type)) { | |
2933 | error_setg(&local_err, "CPU core type should be %s", base_core_type); | |
2934 | goto out; | |
2935 | } | |
2936 | ||
2937 | if (cc->core_id % smp_threads) { | |
2938 | error_setg(&local_err, "invalid core id %d", cc->core_id); | |
2939 | goto out; | |
2940 | } | |
2941 | ||
459264ef DG |
2942 | /* |
2943 | * In general we should have homogeneous threads-per-core, but old | |
2944 | * (pre hotplug support) machine types allow the last core to have | |
2945 | * reduced threads as a compatibility hack for when we allowed | |
2946 | * total vcpus not a multiple of threads-per-core. | |
2947 | */ | |
2948 | if (mc->has_hotpluggable_cpus && (cc->nr_threads != smp_threads)) { | |
8149e299 DG |
2949 | error_setg(errp, "invalid nr-threads %d, must be %d", |
2950 | cc->nr_threads, smp_threads); | |
2951 | return; | |
2952 | } | |
2953 | ||
535455fd IM |
2954 | core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); |
2955 | if (!core_slot) { | |
ff9006dd IM |
2956 | error_setg(&local_err, "core id %d out of range", cc->core_id); |
2957 | goto out; | |
2958 | } | |
2959 | ||
535455fd | 2960 | if (core_slot->cpu) { |
ff9006dd IM |
2961 | error_setg(&local_err, "core %d already populated", cc->core_id); |
2962 | goto out; | |
2963 | } | |
2964 | ||
a0ceb640 | 2965 | numa_cpu_pre_plug(core_slot, dev, &local_err); |
0b8497f0 | 2966 | |
ff9006dd IM |
2967 | out: |
2968 | g_free(base_core_type); | |
2969 | error_propagate(errp, local_err); | |
2970 | } | |
2971 | ||
c20d332a BR |
2972 | static void spapr_machine_device_plug(HotplugHandler *hotplug_dev, |
2973 | DeviceState *dev, Error **errp) | |
2974 | { | |
2975 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine()); | |
2976 | ||
2977 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { | |
b556854b | 2978 | int node; |
c20d332a BR |
2979 | |
2980 | if (!smc->dr_lmb_enabled) { | |
2981 | error_setg(errp, "Memory hotplug not supported for this machine"); | |
2982 | return; | |
2983 | } | |
2984 | node = object_property_get_int(OBJECT(dev), PC_DIMM_NODE_PROP, errp); | |
2985 | if (*errp) { | |
2986 | return; | |
2987 | } | |
1a5512bb GA |
2988 | if (node < 0 || node >= MAX_NODES) { |
2989 | error_setg(errp, "Invaild node %d", node); | |
2990 | return; | |
2991 | } | |
c20d332a | 2992 | |
b556854b BR |
2993 | /* |
2994 | * Currently PowerPC kernel doesn't allow hot-adding memory to | |
2995 | * memory-less node, but instead will silently add the memory | |
2996 | * to the first node that has some memory. This causes two | |
2997 | * unexpected behaviours for the user. | |
2998 | * | |
2999 | * - Memory gets hotplugged to a different node than what the user | |
3000 | * specified. | |
3001 | * - Since pc-dimm subsystem in QEMU still thinks that memory belongs | |
3002 | * to memory-less node, a reboot will set things accordingly | |
3003 | * and the previously hotplugged memory now ends in the right node. | |
3004 | * This appears as if some memory moved from one node to another. | |
3005 | * | |
3006 | * So until kernel starts supporting memory hotplug to memory-less | |
3007 | * nodes, just prevent such attempts upfront in QEMU. | |
3008 | */ | |
3009 | if (nb_numa_nodes && !numa_info[node].node_mem) { | |
3010 | error_setg(errp, "Can't hotplug memory to memory-less node %d", | |
3011 | node); | |
3012 | return; | |
3013 | } | |
3014 | ||
c20d332a | 3015 | spapr_memory_plug(hotplug_dev, dev, node, errp); |
af81cf32 BR |
3016 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { |
3017 | spapr_core_plug(hotplug_dev, dev, errp); | |
c20d332a BR |
3018 | } |
3019 | } | |
3020 | ||
3021 | static void spapr_machine_device_unplug(HotplugHandler *hotplug_dev, | |
3022 | DeviceState *dev, Error **errp) | |
3023 | { | |
cf632463 | 3024 | sPAPRMachineState *sms = SPAPR_MACHINE(qdev_get_machine()); |
3c0c47e3 | 3025 | MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine()); |
6f4b5c3e | 3026 | |
c20d332a | 3027 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { |
cf632463 BR |
3028 | if (spapr_ovec_test(sms->ov5_cas, OV5_HP_EVT)) { |
3029 | spapr_memory_unplug(hotplug_dev, dev, errp); | |
3030 | } else { | |
3031 | error_setg(errp, "Memory hot unplug not supported for this guest"); | |
3032 | } | |
3033 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { | |
c5514d0e | 3034 | if (!mc->has_hotpluggable_cpus) { |
cf632463 BR |
3035 | error_setg(errp, "CPU hot unplug not supported on this machine"); |
3036 | return; | |
3037 | } | |
3038 | spapr_core_unplug(hotplug_dev, dev, errp); | |
3039 | } | |
3040 | } | |
3041 | ||
3042 | static void spapr_machine_device_unplug_request(HotplugHandler *hotplug_dev, | |
3043 | DeviceState *dev, Error **errp) | |
3044 | { | |
3045 | sPAPRMachineState *sms = SPAPR_MACHINE(qdev_get_machine()); | |
3046 | MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine()); | |
3047 | ||
3048 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { | |
3049 | if (spapr_ovec_test(sms->ov5_cas, OV5_HP_EVT)) { | |
3050 | spapr_memory_unplug_request(hotplug_dev, dev, errp); | |
3051 | } else { | |
3052 | /* NOTE: this means there is a window after guest reset, prior to | |
3053 | * CAS negotiation, where unplug requests will fail due to the | |
3054 | * capability not being detected yet. This is a bit different than | |
3055 | * the case with PCI unplug, where the events will be queued and | |
3056 | * eventually handled by the guest after boot | |
3057 | */ | |
3058 | error_setg(errp, "Memory hot unplug not supported for this guest"); | |
3059 | } | |
6f4b5c3e | 3060 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { |
c5514d0e | 3061 | if (!mc->has_hotpluggable_cpus) { |
6f4b5c3e BR |
3062 | error_setg(errp, "CPU hot unplug not supported on this machine"); |
3063 | return; | |
3064 | } | |
115debf2 | 3065 | spapr_core_unplug_request(hotplug_dev, dev, errp); |
c20d332a BR |
3066 | } |
3067 | } | |
3068 | ||
94a94e4c BR |
3069 | static void spapr_machine_device_pre_plug(HotplugHandler *hotplug_dev, |
3070 | DeviceState *dev, Error **errp) | |
3071 | { | |
c871bc70 LV |
3072 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { |
3073 | spapr_memory_pre_plug(hotplug_dev, dev, errp); | |
3074 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { | |
94a94e4c BR |
3075 | spapr_core_pre_plug(hotplug_dev, dev, errp); |
3076 | } | |
3077 | } | |
3078 | ||
7ebaf795 BR |
3079 | static HotplugHandler *spapr_get_hotplug_handler(MachineState *machine, |
3080 | DeviceState *dev) | |
c20d332a | 3081 | { |
94a94e4c BR |
3082 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) || |
3083 | object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { | |
c20d332a BR |
3084 | return HOTPLUG_HANDLER(machine); |
3085 | } | |
3086 | return NULL; | |
3087 | } | |
3088 | ||
ea089eeb IM |
3089 | static CpuInstanceProperties |
3090 | spapr_cpu_index_to_props(MachineState *machine, unsigned cpu_index) | |
20bb648d | 3091 | { |
ea089eeb IM |
3092 | CPUArchId *core_slot; |
3093 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
3094 | ||
3095 | /* make sure possible_cpu are intialized */ | |
3096 | mc->possible_cpu_arch_ids(machine); | |
3097 | /* get CPU core slot containing thread that matches cpu_index */ | |
3098 | core_slot = spapr_find_cpu_slot(machine, cpu_index, NULL); | |
3099 | assert(core_slot); | |
3100 | return core_slot->props; | |
20bb648d DG |
3101 | } |
3102 | ||
535455fd IM |
3103 | static const CPUArchIdList *spapr_possible_cpu_arch_ids(MachineState *machine) |
3104 | { | |
3105 | int i; | |
3106 | int spapr_max_cores = max_cpus / smp_threads; | |
3107 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
3108 | ||
c5514d0e | 3109 | if (!mc->has_hotpluggable_cpus) { |
535455fd IM |
3110 | spapr_max_cores = QEMU_ALIGN_UP(smp_cpus, smp_threads) / smp_threads; |
3111 | } | |
3112 | if (machine->possible_cpus) { | |
3113 | assert(machine->possible_cpus->len == spapr_max_cores); | |
3114 | return machine->possible_cpus; | |
3115 | } | |
3116 | ||
3117 | machine->possible_cpus = g_malloc0(sizeof(CPUArchIdList) + | |
3118 | sizeof(CPUArchId) * spapr_max_cores); | |
3119 | machine->possible_cpus->len = spapr_max_cores; | |
3120 | for (i = 0; i < machine->possible_cpus->len; i++) { | |
3121 | int core_id = i * smp_threads; | |
3122 | ||
f2d672c2 | 3123 | machine->possible_cpus->cpus[i].vcpus_count = smp_threads; |
535455fd IM |
3124 | machine->possible_cpus->cpus[i].arch_id = core_id; |
3125 | machine->possible_cpus->cpus[i].props.has_core_id = true; | |
3126 | machine->possible_cpus->cpus[i].props.core_id = core_id; | |
ea089eeb IM |
3127 | |
3128 | /* default distribution of CPUs over NUMA nodes */ | |
3129 | if (nb_numa_nodes) { | |
3130 | /* preset values but do not enable them i.e. 'has_node_id = false', | |
3131 | * numa init code will enable them later if manual mapping wasn't | |
3132 | * present on CLI */ | |
3133 | machine->possible_cpus->cpus[i].props.node_id = | |
3134 | core_id / smp_threads / smp_cores % nb_numa_nodes; | |
3135 | } | |
535455fd IM |
3136 | } |
3137 | return machine->possible_cpus; | |
3138 | } | |
3139 | ||
6737d9ad | 3140 | static void spapr_phb_placement(sPAPRMachineState *spapr, uint32_t index, |
daa23699 DG |
3141 | uint64_t *buid, hwaddr *pio, |
3142 | hwaddr *mmio32, hwaddr *mmio64, | |
6737d9ad DG |
3143 | unsigned n_dma, uint32_t *liobns, Error **errp) |
3144 | { | |
357d1e3b DG |
3145 | /* |
3146 | * New-style PHB window placement. | |
3147 | * | |
3148 | * Goals: Gives large (1TiB), naturally aligned 64-bit MMIO window | |
3149 | * for each PHB, in addition to 2GiB 32-bit MMIO and 64kiB PIO | |
3150 | * windows. | |
3151 | * | |
3152 | * Some guest kernels can't work with MMIO windows above 1<<46 | |
3153 | * (64TiB), so we place up to 31 PHBs in the area 32TiB..64TiB | |
3154 | * | |
3155 | * 32TiB..(33TiB+1984kiB) contains the 64kiB PIO windows for each | |
3156 | * PHB stacked together. (32TiB+2GiB)..(32TiB+64GiB) contains the | |
3157 | * 2GiB 32-bit MMIO windows for each PHB. Then 33..64TiB has the | |
3158 | * 1TiB 64-bit MMIO windows for each PHB. | |
3159 | */ | |
6737d9ad | 3160 | const uint64_t base_buid = 0x800000020000000ULL; |
25e6a118 MT |
3161 | #define SPAPR_MAX_PHBS ((SPAPR_PCI_LIMIT - SPAPR_PCI_BASE) / \ |
3162 | SPAPR_PCI_MEM64_WIN_SIZE - 1) | |
6737d9ad DG |
3163 | int i; |
3164 | ||
357d1e3b DG |
3165 | /* Sanity check natural alignments */ |
3166 | QEMU_BUILD_BUG_ON((SPAPR_PCI_BASE % SPAPR_PCI_MEM64_WIN_SIZE) != 0); | |
3167 | QEMU_BUILD_BUG_ON((SPAPR_PCI_LIMIT % SPAPR_PCI_MEM64_WIN_SIZE) != 0); | |
3168 | QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM64_WIN_SIZE % SPAPR_PCI_MEM32_WIN_SIZE) != 0); | |
3169 | QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM32_WIN_SIZE % SPAPR_PCI_IO_WIN_SIZE) != 0); | |
3170 | /* Sanity check bounds */ | |
25e6a118 MT |
3171 | QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_IO_WIN_SIZE) > |
3172 | SPAPR_PCI_MEM32_WIN_SIZE); | |
3173 | QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_MEM32_WIN_SIZE) > | |
3174 | SPAPR_PCI_MEM64_WIN_SIZE); | |
3175 | ||
3176 | if (index >= SPAPR_MAX_PHBS) { | |
3177 | error_setg(errp, "\"index\" for PAPR PHB is too large (max %llu)", | |
3178 | SPAPR_MAX_PHBS - 1); | |
6737d9ad DG |
3179 | return; |
3180 | } | |
3181 | ||
3182 | *buid = base_buid + index; | |
3183 | for (i = 0; i < n_dma; ++i) { | |
3184 | liobns[i] = SPAPR_PCI_LIOBN(index, i); | |
3185 | } | |
3186 | ||
357d1e3b DG |
3187 | *pio = SPAPR_PCI_BASE + index * SPAPR_PCI_IO_WIN_SIZE; |
3188 | *mmio32 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM32_WIN_SIZE; | |
3189 | *mmio64 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM64_WIN_SIZE; | |
6737d9ad DG |
3190 | } |
3191 | ||
7844e12b CLG |
3192 | static ICSState *spapr_ics_get(XICSFabric *dev, int irq) |
3193 | { | |
3194 | sPAPRMachineState *spapr = SPAPR_MACHINE(dev); | |
3195 | ||
3196 | return ics_valid_irq(spapr->ics, irq) ? spapr->ics : NULL; | |
3197 | } | |
3198 | ||
3199 | static void spapr_ics_resend(XICSFabric *dev) | |
3200 | { | |
3201 | sPAPRMachineState *spapr = SPAPR_MACHINE(dev); | |
3202 | ||
3203 | ics_resend(spapr->ics); | |
3204 | } | |
3205 | ||
06747ba6 | 3206 | static ICPState *spapr_icp_get(XICSFabric *xi, int cpu_dt_id) |
b2fc59aa | 3207 | { |
5bc8d26d | 3208 | PowerPCCPU *cpu = ppc_get_vcpu_by_dt_id(cpu_dt_id); |
b2fc59aa | 3209 | |
5bc8d26d | 3210 | return cpu ? ICP(cpu->intc) : NULL; |
b2fc59aa CLG |
3211 | } |
3212 | ||
6449da45 CLG |
3213 | static void spapr_pic_print_info(InterruptStatsProvider *obj, |
3214 | Monitor *mon) | |
3215 | { | |
3216 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
5bc8d26d CLG |
3217 | CPUState *cs; |
3218 | ||
3219 | CPU_FOREACH(cs) { | |
3220 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
6449da45 | 3221 | |
5bc8d26d | 3222 | icp_pic_print_info(ICP(cpu->intc), mon); |
6449da45 CLG |
3223 | } |
3224 | ||
3225 | ics_pic_print_info(spapr->ics, mon); | |
3226 | } | |
3227 | ||
29ee3247 AK |
3228 | static void spapr_machine_class_init(ObjectClass *oc, void *data) |
3229 | { | |
3230 | MachineClass *mc = MACHINE_CLASS(oc); | |
224245bf | 3231 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc); |
71461b0f | 3232 | FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc); |
34316482 | 3233 | NMIClass *nc = NMI_CLASS(oc); |
c20d332a | 3234 | HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc); |
1d1be34d | 3235 | PPCVirtualHypervisorClass *vhc = PPC_VIRTUAL_HYPERVISOR_CLASS(oc); |
7844e12b | 3236 | XICSFabricClass *xic = XICS_FABRIC_CLASS(oc); |
6449da45 | 3237 | InterruptStatsProviderClass *ispc = INTERRUPT_STATS_PROVIDER_CLASS(oc); |
958db90c | 3238 | |
0eb9054c | 3239 | mc->desc = "pSeries Logical Partition (PAPR compliant)"; |
fc9f38c3 DG |
3240 | |
3241 | /* | |
3242 | * We set up the default / latest behaviour here. The class_init | |
3243 | * functions for the specific versioned machine types can override | |
3244 | * these details for backwards compatibility | |
3245 | */ | |
958db90c MA |
3246 | mc->init = ppc_spapr_init; |
3247 | mc->reset = ppc_spapr_reset; | |
3248 | mc->block_default_type = IF_SCSI; | |
6244bb7e | 3249 | mc->max_cpus = 1024; |
958db90c | 3250 | mc->no_parallel = 1; |
5b2128d2 | 3251 | mc->default_boot_order = ""; |
a34944fe | 3252 | mc->default_ram_size = 512 * M_BYTE; |
958db90c | 3253 | mc->kvm_type = spapr_kvm_type; |
9e3f9733 | 3254 | mc->has_dynamic_sysbus = true; |
e4024630 | 3255 | mc->pci_allow_0_address = true; |
7ebaf795 | 3256 | mc->get_hotplug_handler = spapr_get_hotplug_handler; |
94a94e4c | 3257 | hc->pre_plug = spapr_machine_device_pre_plug; |
c20d332a BR |
3258 | hc->plug = spapr_machine_device_plug; |
3259 | hc->unplug = spapr_machine_device_unplug; | |
ea089eeb | 3260 | mc->cpu_index_to_instance_props = spapr_cpu_index_to_props; |
535455fd | 3261 | mc->possible_cpu_arch_ids = spapr_possible_cpu_arch_ids; |
cf632463 | 3262 | hc->unplug_request = spapr_machine_device_unplug_request; |
00b4fbe2 | 3263 | |
fc9f38c3 | 3264 | smc->dr_lmb_enabled = true; |
3daa4a9f | 3265 | smc->tcg_default_cpu = "POWER8"; |
c5514d0e | 3266 | mc->has_hotpluggable_cpus = true; |
71461b0f | 3267 | fwc->get_dev_path = spapr_get_fw_dev_path; |
34316482 | 3268 | nc->nmi_monitor_handler = spapr_nmi; |
6737d9ad | 3269 | smc->phb_placement = spapr_phb_placement; |
1d1be34d | 3270 | vhc->hypercall = emulate_spapr_hypercall; |
e57ca75c DG |
3271 | vhc->hpt_mask = spapr_hpt_mask; |
3272 | vhc->map_hptes = spapr_map_hptes; | |
3273 | vhc->unmap_hptes = spapr_unmap_hptes; | |
3274 | vhc->store_hpte = spapr_store_hpte; | |
9861bb3e | 3275 | vhc->get_patbe = spapr_get_patbe; |
7844e12b CLG |
3276 | xic->ics_get = spapr_ics_get; |
3277 | xic->ics_resend = spapr_ics_resend; | |
b2fc59aa | 3278 | xic->icp_get = spapr_icp_get; |
6449da45 | 3279 | ispc->print_info = spapr_pic_print_info; |
55641213 LV |
3280 | /* Force NUMA node memory size to be a multiple of |
3281 | * SPAPR_MEMORY_BLOCK_SIZE (256M) since that's the granularity | |
3282 | * in which LMBs are represented and hot-added | |
3283 | */ | |
3284 | mc->numa_mem_align_shift = 28; | |
29ee3247 AK |
3285 | } |
3286 | ||
3287 | static const TypeInfo spapr_machine_info = { | |
3288 | .name = TYPE_SPAPR_MACHINE, | |
3289 | .parent = TYPE_MACHINE, | |
4aee7362 | 3290 | .abstract = true, |
6ca1502e | 3291 | .instance_size = sizeof(sPAPRMachineState), |
23825581 | 3292 | .instance_init = spapr_machine_initfn, |
87bbdd9c | 3293 | .instance_finalize = spapr_machine_finalizefn, |
183930c0 | 3294 | .class_size = sizeof(sPAPRMachineClass), |
29ee3247 | 3295 | .class_init = spapr_machine_class_init, |
71461b0f AK |
3296 | .interfaces = (InterfaceInfo[]) { |
3297 | { TYPE_FW_PATH_PROVIDER }, | |
34316482 | 3298 | { TYPE_NMI }, |
c20d332a | 3299 | { TYPE_HOTPLUG_HANDLER }, |
1d1be34d | 3300 | { TYPE_PPC_VIRTUAL_HYPERVISOR }, |
7844e12b | 3301 | { TYPE_XICS_FABRIC }, |
6449da45 | 3302 | { TYPE_INTERRUPT_STATS_PROVIDER }, |
71461b0f AK |
3303 | { } |
3304 | }, | |
29ee3247 AK |
3305 | }; |
3306 | ||
fccbc785 | 3307 | #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest) \ |
5013c547 DG |
3308 | static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \ |
3309 | void *data) \ | |
3310 | { \ | |
3311 | MachineClass *mc = MACHINE_CLASS(oc); \ | |
3312 | spapr_machine_##suffix##_class_options(mc); \ | |
fccbc785 DG |
3313 | if (latest) { \ |
3314 | mc->alias = "pseries"; \ | |
3315 | mc->is_default = 1; \ | |
3316 | } \ | |
5013c547 DG |
3317 | } \ |
3318 | static void spapr_machine_##suffix##_instance_init(Object *obj) \ | |
3319 | { \ | |
3320 | MachineState *machine = MACHINE(obj); \ | |
3321 | spapr_machine_##suffix##_instance_options(machine); \ | |
3322 | } \ | |
3323 | static const TypeInfo spapr_machine_##suffix##_info = { \ | |
3324 | .name = MACHINE_TYPE_NAME("pseries-" verstr), \ | |
3325 | .parent = TYPE_SPAPR_MACHINE, \ | |
3326 | .class_init = spapr_machine_##suffix##_class_init, \ | |
3327 | .instance_init = spapr_machine_##suffix##_instance_init, \ | |
3328 | }; \ | |
3329 | static void spapr_machine_register_##suffix(void) \ | |
3330 | { \ | |
3331 | type_register(&spapr_machine_##suffix##_info); \ | |
3332 | } \ | |
0e6aac87 | 3333 | type_init(spapr_machine_register_##suffix) |
5013c547 | 3334 | |
3fa14fbe DG |
3335 | /* |
3336 | * pseries-2.10 | |
3337 | */ | |
3338 | static void spapr_machine_2_10_instance_options(MachineState *machine) | |
3339 | { | |
3340 | } | |
3341 | ||
3342 | static void spapr_machine_2_10_class_options(MachineClass *mc) | |
3343 | { | |
3344 | /* Defaults for the latest behaviour inherited from the base class */ | |
3345 | } | |
3346 | ||
3347 | DEFINE_SPAPR_MACHINE(2_10, "2.10", true); | |
3348 | ||
fa325e6c DG |
3349 | /* |
3350 | * pseries-2.9 | |
3351 | */ | |
3fa14fbe DG |
3352 | #define SPAPR_COMPAT_2_9 \ |
3353 | HW_COMPAT_2_9 | |
3354 | ||
fa325e6c DG |
3355 | static void spapr_machine_2_9_instance_options(MachineState *machine) |
3356 | { | |
3fa14fbe | 3357 | spapr_machine_2_10_instance_options(machine); |
fa325e6c DG |
3358 | } |
3359 | ||
3360 | static void spapr_machine_2_9_class_options(MachineClass *mc) | |
3361 | { | |
3fa14fbe DG |
3362 | spapr_machine_2_10_class_options(mc); |
3363 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_9); | |
3bfe5716 | 3364 | mc->numa_auto_assign_ram = numa_legacy_auto_assign_ram; |
fa325e6c DG |
3365 | } |
3366 | ||
3fa14fbe | 3367 | DEFINE_SPAPR_MACHINE(2_9, "2.9", false); |
fa325e6c | 3368 | |
db800b21 DG |
3369 | /* |
3370 | * pseries-2.8 | |
3371 | */ | |
82516263 DG |
3372 | #define SPAPR_COMPAT_2_8 \ |
3373 | HW_COMPAT_2_8 \ | |
3374 | { \ | |
3375 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3376 | .property = "pcie-extended-configuration-space", \ | |
3377 | .value = "off", \ | |
3378 | }, | |
fa325e6c | 3379 | |
db800b21 DG |
3380 | static void spapr_machine_2_8_instance_options(MachineState *machine) |
3381 | { | |
fa325e6c | 3382 | spapr_machine_2_9_instance_options(machine); |
db800b21 DG |
3383 | } |
3384 | ||
3385 | static void spapr_machine_2_8_class_options(MachineClass *mc) | |
3386 | { | |
fa325e6c DG |
3387 | spapr_machine_2_9_class_options(mc); |
3388 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8); | |
55641213 | 3389 | mc->numa_mem_align_shift = 23; |
db800b21 DG |
3390 | } |
3391 | ||
fa325e6c | 3392 | DEFINE_SPAPR_MACHINE(2_8, "2.8", false); |
db800b21 | 3393 | |
1ea1eefc BR |
3394 | /* |
3395 | * pseries-2.7 | |
3396 | */ | |
357d1e3b DG |
3397 | #define SPAPR_COMPAT_2_7 \ |
3398 | HW_COMPAT_2_7 \ | |
3399 | { \ | |
3400 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3401 | .property = "mem_win_size", \ | |
3402 | .value = stringify(SPAPR_PCI_2_7_MMIO_WIN_SIZE),\ | |
3403 | }, \ | |
3404 | { \ | |
3405 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3406 | .property = "mem64_win_size", \ | |
3407 | .value = "0", \ | |
146c11f1 DG |
3408 | }, \ |
3409 | { \ | |
3410 | .driver = TYPE_POWERPC_CPU, \ | |
3411 | .property = "pre-2.8-migration", \ | |
3412 | .value = "on", \ | |
5c4537bd DG |
3413 | }, \ |
3414 | { \ | |
3415 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3416 | .property = "pre-2.8-migration", \ | |
3417 | .value = "on", \ | |
357d1e3b DG |
3418 | }, |
3419 | ||
3420 | static void phb_placement_2_7(sPAPRMachineState *spapr, uint32_t index, | |
3421 | uint64_t *buid, hwaddr *pio, | |
3422 | hwaddr *mmio32, hwaddr *mmio64, | |
3423 | unsigned n_dma, uint32_t *liobns, Error **errp) | |
3424 | { | |
3425 | /* Legacy PHB placement for pseries-2.7 and earlier machine types */ | |
3426 | const uint64_t base_buid = 0x800000020000000ULL; | |
3427 | const hwaddr phb_spacing = 0x1000000000ULL; /* 64 GiB */ | |
3428 | const hwaddr mmio_offset = 0xa0000000; /* 2 GiB + 512 MiB */ | |
3429 | const hwaddr pio_offset = 0x80000000; /* 2 GiB */ | |
3430 | const uint32_t max_index = 255; | |
3431 | const hwaddr phb0_alignment = 0x10000000000ULL; /* 1 TiB */ | |
3432 | ||
3433 | uint64_t ram_top = MACHINE(spapr)->ram_size; | |
3434 | hwaddr phb0_base, phb_base; | |
3435 | int i; | |
3436 | ||
3437 | /* Do we have hotpluggable memory? */ | |
3438 | if (MACHINE(spapr)->maxram_size > ram_top) { | |
3439 | /* Can't just use maxram_size, because there may be an | |
3440 | * alignment gap between normal and hotpluggable memory | |
3441 | * regions */ | |
3442 | ram_top = spapr->hotplug_memory.base + | |
3443 | memory_region_size(&spapr->hotplug_memory.mr); | |
3444 | } | |
3445 | ||
3446 | phb0_base = QEMU_ALIGN_UP(ram_top, phb0_alignment); | |
3447 | ||
3448 | if (index > max_index) { | |
3449 | error_setg(errp, "\"index\" for PAPR PHB is too large (max %u)", | |
3450 | max_index); | |
3451 | return; | |
3452 | } | |
3453 | ||
3454 | *buid = base_buid + index; | |
3455 | for (i = 0; i < n_dma; ++i) { | |
3456 | liobns[i] = SPAPR_PCI_LIOBN(index, i); | |
3457 | } | |
3458 | ||
3459 | phb_base = phb0_base + index * phb_spacing; | |
3460 | *pio = phb_base + pio_offset; | |
3461 | *mmio32 = phb_base + mmio_offset; | |
3462 | /* | |
3463 | * We don't set the 64-bit MMIO window, relying on the PHB's | |
3464 | * fallback behaviour of automatically splitting a large "32-bit" | |
3465 | * window into contiguous 32-bit and 64-bit windows | |
3466 | */ | |
3467 | } | |
db800b21 | 3468 | |
1ea1eefc BR |
3469 | static void spapr_machine_2_7_instance_options(MachineState *machine) |
3470 | { | |
f6229214 MR |
3471 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
3472 | ||
672de881 | 3473 | spapr_machine_2_8_instance_options(machine); |
f6229214 | 3474 | spapr->use_hotplug_event_source = false; |
1ea1eefc BR |
3475 | } |
3476 | ||
3477 | static void spapr_machine_2_7_class_options(MachineClass *mc) | |
3478 | { | |
3daa4a9f TH |
3479 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3480 | ||
db800b21 | 3481 | spapr_machine_2_8_class_options(mc); |
3daa4a9f | 3482 | smc->tcg_default_cpu = "POWER7"; |
db800b21 | 3483 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_7); |
357d1e3b | 3484 | smc->phb_placement = phb_placement_2_7; |
1ea1eefc BR |
3485 | } |
3486 | ||
db800b21 | 3487 | DEFINE_SPAPR_MACHINE(2_7, "2.7", false); |
1ea1eefc | 3488 | |
4b23699c DG |
3489 | /* |
3490 | * pseries-2.6 | |
3491 | */ | |
1ea1eefc | 3492 | #define SPAPR_COMPAT_2_6 \ |
ae4de14c AK |
3493 | HW_COMPAT_2_6 \ |
3494 | { \ | |
3495 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ | |
3496 | .property = "ddw",\ | |
3497 | .value = stringify(off),\ | |
3498 | }, | |
1ea1eefc | 3499 | |
4b23699c DG |
3500 | static void spapr_machine_2_6_instance_options(MachineState *machine) |
3501 | { | |
672de881 | 3502 | spapr_machine_2_7_instance_options(machine); |
4b23699c DG |
3503 | } |
3504 | ||
3505 | static void spapr_machine_2_6_class_options(MachineClass *mc) | |
3506 | { | |
1ea1eefc | 3507 | spapr_machine_2_7_class_options(mc); |
c5514d0e | 3508 | mc->has_hotpluggable_cpus = false; |
1ea1eefc | 3509 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_6); |
4b23699c DG |
3510 | } |
3511 | ||
1ea1eefc | 3512 | DEFINE_SPAPR_MACHINE(2_6, "2.6", false); |
4b23699c | 3513 | |
1c5f29bb DG |
3514 | /* |
3515 | * pseries-2.5 | |
3516 | */ | |
4b23699c | 3517 | #define SPAPR_COMPAT_2_5 \ |
57c522f4 TH |
3518 | HW_COMPAT_2_5 \ |
3519 | { \ | |
3520 | .driver = "spapr-vlan", \ | |
3521 | .property = "use-rx-buffer-pools", \ | |
3522 | .value = "off", \ | |
3523 | }, | |
4b23699c | 3524 | |
5013c547 | 3525 | static void spapr_machine_2_5_instance_options(MachineState *machine) |
1c5f29bb | 3526 | { |
672de881 | 3527 | spapr_machine_2_6_instance_options(machine); |
5013c547 DG |
3528 | } |
3529 | ||
3530 | static void spapr_machine_2_5_class_options(MachineClass *mc) | |
3531 | { | |
57040d45 TH |
3532 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3533 | ||
4b23699c | 3534 | spapr_machine_2_6_class_options(mc); |
57040d45 | 3535 | smc->use_ohci_by_default = true; |
4b23699c | 3536 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5); |
1c5f29bb DG |
3537 | } |
3538 | ||
4b23699c | 3539 | DEFINE_SPAPR_MACHINE(2_5, "2.5", false); |
1c5f29bb DG |
3540 | |
3541 | /* | |
3542 | * pseries-2.4 | |
3543 | */ | |
80fd50f9 CH |
3544 | #define SPAPR_COMPAT_2_4 \ |
3545 | HW_COMPAT_2_4 | |
3546 | ||
5013c547 | 3547 | static void spapr_machine_2_4_instance_options(MachineState *machine) |
1c5f29bb | 3548 | { |
5013c547 DG |
3549 | spapr_machine_2_5_instance_options(machine); |
3550 | } | |
1c5f29bb | 3551 | |
5013c547 DG |
3552 | static void spapr_machine_2_4_class_options(MachineClass *mc) |
3553 | { | |
fc9f38c3 DG |
3554 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3555 | ||
3556 | spapr_machine_2_5_class_options(mc); | |
fc9f38c3 | 3557 | smc->dr_lmb_enabled = false; |
f949b4e5 | 3558 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4); |
1c5f29bb DG |
3559 | } |
3560 | ||
fccbc785 | 3561 | DEFINE_SPAPR_MACHINE(2_4, "2.4", false); |
1c5f29bb DG |
3562 | |
3563 | /* | |
3564 | * pseries-2.3 | |
3565 | */ | |
38ff32c6 | 3566 | #define SPAPR_COMPAT_2_3 \ |
7619c7b0 MR |
3567 | HW_COMPAT_2_3 \ |
3568 | {\ | |
3569 | .driver = "spapr-pci-host-bridge",\ | |
3570 | .property = "dynamic-reconfiguration",\ | |
3571 | .value = "off",\ | |
3572 | }, | |
38ff32c6 | 3573 | |
5013c547 | 3574 | static void spapr_machine_2_3_instance_options(MachineState *machine) |
d25228e7 | 3575 | { |
5013c547 | 3576 | spapr_machine_2_4_instance_options(machine); |
ff14e817 | 3577 | savevm_skip_section_footers(); |
13d16814 | 3578 | global_state_set_optional(); |
09b5e30d | 3579 | savevm_skip_configuration(); |
d25228e7 JW |
3580 | } |
3581 | ||
5013c547 | 3582 | static void spapr_machine_2_3_class_options(MachineClass *mc) |
6026db45 | 3583 | { |
fc9f38c3 | 3584 | spapr_machine_2_4_class_options(mc); |
f949b4e5 | 3585 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3); |
6026db45 | 3586 | } |
fccbc785 | 3587 | DEFINE_SPAPR_MACHINE(2_3, "2.3", false); |
6026db45 | 3588 | |
1c5f29bb DG |
3589 | /* |
3590 | * pseries-2.2 | |
3591 | */ | |
3592 | ||
3593 | #define SPAPR_COMPAT_2_2 \ | |
1c5f29bb DG |
3594 | HW_COMPAT_2_2 \ |
3595 | {\ | |
3596 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ | |
3597 | .property = "mem_win_size",\ | |
3598 | .value = "0x20000000",\ | |
3599 | }, | |
3600 | ||
5013c547 | 3601 | static void spapr_machine_2_2_instance_options(MachineState *machine) |
1c5f29bb | 3602 | { |
5013c547 | 3603 | spapr_machine_2_3_instance_options(machine); |
cba0e779 | 3604 | machine->suppress_vmdesc = true; |
1c5f29bb DG |
3605 | } |
3606 | ||
5013c547 | 3607 | static void spapr_machine_2_2_class_options(MachineClass *mc) |
4aee7362 | 3608 | { |
fc9f38c3 | 3609 | spapr_machine_2_3_class_options(mc); |
f949b4e5 | 3610 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2); |
4aee7362 | 3611 | } |
fccbc785 | 3612 | DEFINE_SPAPR_MACHINE(2_2, "2.2", false); |
4aee7362 | 3613 | |
1c5f29bb DG |
3614 | /* |
3615 | * pseries-2.1 | |
3616 | */ | |
3617 | #define SPAPR_COMPAT_2_1 \ | |
1c5f29bb | 3618 | HW_COMPAT_2_1 |
3dab0244 | 3619 | |
5013c547 | 3620 | static void spapr_machine_2_1_instance_options(MachineState *machine) |
1c5f29bb | 3621 | { |
5013c547 | 3622 | spapr_machine_2_2_instance_options(machine); |
1c5f29bb | 3623 | } |
d25228e7 | 3624 | |
5013c547 | 3625 | static void spapr_machine_2_1_class_options(MachineClass *mc) |
d25228e7 | 3626 | { |
fc9f38c3 | 3627 | spapr_machine_2_2_class_options(mc); |
f949b4e5 | 3628 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1); |
d25228e7 | 3629 | } |
fccbc785 | 3630 | DEFINE_SPAPR_MACHINE(2_1, "2.1", false); |
fb0fc8f6 | 3631 | |
29ee3247 | 3632 | static void spapr_machine_register_types(void) |
9fdf0c29 | 3633 | { |
29ee3247 | 3634 | type_register_static(&spapr_machine_info); |
9fdf0c29 DG |
3635 | } |
3636 | ||
29ee3247 | 3637 | type_init(spapr_machine_register_types) |