]>
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" |
c4b63b7c | 41 | #include "migration/misc.h" |
84a899de | 42 | #include "migration/global_state.h" |
f2a8f0a6 | 43 | #include "migration/register.h" |
4be21d56 | 44 | #include "mmu-hash64.h" |
b4db5413 | 45 | #include "mmu-book3s-v3.h" |
3794d548 | 46 | #include "qom/cpu.h" |
9fdf0c29 DG |
47 | |
48 | #include "hw/boards.h" | |
0d09e41a | 49 | #include "hw/ppc/ppc.h" |
9fdf0c29 DG |
50 | #include "hw/loader.h" |
51 | ||
7804c353 | 52 | #include "hw/ppc/fdt.h" |
0d09e41a PB |
53 | #include "hw/ppc/spapr.h" |
54 | #include "hw/ppc/spapr_vio.h" | |
55 | #include "hw/pci-host/spapr.h" | |
56 | #include "hw/ppc/xics.h" | |
a2cb15b0 | 57 | #include "hw/pci/msi.h" |
9fdf0c29 | 58 | |
83c9f4ca | 59 | #include "hw/pci/pci.h" |
71461b0f AK |
60 | #include "hw/scsi/scsi.h" |
61 | #include "hw/virtio/virtio-scsi.h" | |
c4e13492 | 62 | #include "hw/virtio/vhost-scsi-common.h" |
f61b4bed | 63 | |
022c62cb | 64 | #include "exec/address-spaces.h" |
35139a59 | 65 | #include "hw/usb.h" |
1de7afc9 | 66 | #include "qemu/config-file.h" |
135a129a | 67 | #include "qemu/error-report.h" |
2a6593cb | 68 | #include "trace.h" |
34316482 | 69 | #include "hw/nmi.h" |
6449da45 | 70 | #include "hw/intc/intc.h" |
890c2b77 | 71 | |
68a27b20 | 72 | #include "hw/compat.h" |
f348b6d1 | 73 | #include "qemu/cutils.h" |
94a94e4c | 74 | #include "hw/ppc/spapr_cpu_core.h" |
2474bfd4 | 75 | #include "qmp-commands.h" |
68a27b20 | 76 | |
9fdf0c29 DG |
77 | #include <libfdt.h> |
78 | ||
4d8d5467 BH |
79 | /* SLOF memory layout: |
80 | * | |
81 | * SLOF raw image loaded at 0, copies its romfs right below the flat | |
82 | * device-tree, then position SLOF itself 31M below that | |
83 | * | |
84 | * So we set FW_OVERHEAD to 40MB which should account for all of that | |
85 | * and more | |
86 | * | |
87 | * We load our kernel at 4M, leaving space for SLOF initial image | |
88 | */ | |
38b02bd8 | 89 | #define FDT_MAX_SIZE 0x100000 |
39ac8455 | 90 | #define RTAS_MAX_SIZE 0x10000 |
b7d1f77a | 91 | #define RTAS_MAX_ADDR 0x80000000 /* RTAS must stay below that */ |
a9f8ad8f DG |
92 | #define FW_MAX_SIZE 0x400000 |
93 | #define FW_FILE_NAME "slof.bin" | |
4d8d5467 BH |
94 | #define FW_OVERHEAD 0x2800000 |
95 | #define KERNEL_LOAD_ADDR FW_MAX_SIZE | |
a9f8ad8f | 96 | |
4d8d5467 | 97 | #define MIN_RMA_SLOF 128UL |
9fdf0c29 | 98 | |
0c103f8e DG |
99 | #define PHANDLE_XICP 0x00001111 |
100 | ||
71cd4dac CLG |
101 | static ICSState *spapr_ics_create(sPAPRMachineState *spapr, |
102 | const char *type_ics, | |
103 | int nr_irqs, Error **errp) | |
c04d6cfa | 104 | { |
175d2aa0 | 105 | Error *local_err = NULL; |
71cd4dac | 106 | Object *obj; |
4e4169f7 | 107 | |
71cd4dac | 108 | obj = object_new(type_ics); |
175d2aa0 | 109 | object_property_add_child(OBJECT(spapr), "ics", obj, &error_abort); |
ad265631 GK |
110 | object_property_add_const_link(obj, ICS_PROP_XICS, OBJECT(spapr), |
111 | &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 | ||
46f7afa3 GK |
128 | static bool pre_2_10_vmstate_dummy_icp_needed(void *opaque) |
129 | { | |
130 | /* Dummy entries correspond to unused ICPState objects in older QEMUs, | |
131 | * and newer QEMUs don't even have them. In both cases, we don't want | |
132 | * to send anything on the wire. | |
133 | */ | |
134 | return false; | |
135 | } | |
136 | ||
137 | static const VMStateDescription pre_2_10_vmstate_dummy_icp = { | |
138 | .name = "icp/server", | |
139 | .version_id = 1, | |
140 | .minimum_version_id = 1, | |
141 | .needed = pre_2_10_vmstate_dummy_icp_needed, | |
142 | .fields = (VMStateField[]) { | |
143 | VMSTATE_UNUSED(4), /* uint32_t xirr */ | |
144 | VMSTATE_UNUSED(1), /* uint8_t pending_priority */ | |
145 | VMSTATE_UNUSED(1), /* uint8_t mfrr */ | |
146 | VMSTATE_END_OF_LIST() | |
147 | }, | |
148 | }; | |
149 | ||
150 | static void pre_2_10_vmstate_register_dummy_icp(int i) | |
151 | { | |
152 | vmstate_register(NULL, i, &pre_2_10_vmstate_dummy_icp, | |
153 | (void *)(uintptr_t) i); | |
154 | } | |
155 | ||
156 | static void pre_2_10_vmstate_unregister_dummy_icp(int i) | |
157 | { | |
158 | vmstate_unregister(NULL, &pre_2_10_vmstate_dummy_icp, | |
159 | (void *)(uintptr_t) i); | |
160 | } | |
161 | ||
162 | static inline int xics_max_server_number(void) | |
163 | { | |
164 | return DIV_ROUND_UP(max_cpus * kvmppc_smt_threads(), smp_threads); | |
165 | } | |
166 | ||
71cd4dac | 167 | static void xics_system_init(MachineState *machine, int nr_irqs, Error **errp) |
c04d6cfa | 168 | { |
71cd4dac | 169 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
46f7afa3 | 170 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); |
c04d6cfa | 171 | |
11ad93f6 | 172 | if (kvm_enabled()) { |
2192a930 | 173 | if (machine_kernel_irqchip_allowed(machine) && |
71cd4dac CLG |
174 | !xics_kvm_init(spapr, errp)) { |
175 | spapr->icp_type = TYPE_KVM_ICP; | |
3d85885a | 176 | spapr->ics = spapr_ics_create(spapr, TYPE_ICS_KVM, nr_irqs, errp); |
11ad93f6 | 177 | } |
71cd4dac | 178 | if (machine_kernel_irqchip_required(machine) && !spapr->ics) { |
3d85885a GK |
179 | error_prepend(errp, "kernel_irqchip requested but unavailable: "); |
180 | return; | |
11ad93f6 DG |
181 | } |
182 | } | |
183 | ||
71cd4dac | 184 | if (!spapr->ics) { |
f63ebfe0 | 185 | xics_spapr_init(spapr); |
71cd4dac CLG |
186 | spapr->icp_type = TYPE_ICP; |
187 | spapr->ics = spapr_ics_create(spapr, TYPE_ICS_SIMPLE, nr_irqs, errp); | |
3d85885a GK |
188 | if (!spapr->ics) { |
189 | return; | |
190 | } | |
c04d6cfa | 191 | } |
46f7afa3 GK |
192 | |
193 | if (smc->pre_2_10_has_unused_icps) { | |
194 | int i; | |
195 | ||
196 | for (i = 0; i < xics_max_server_number(); i++) { | |
197 | /* Dummy entries get deregistered when real ICPState objects | |
198 | * are registered during CPU core hotplug. | |
199 | */ | |
200 | pre_2_10_vmstate_register_dummy_icp(i); | |
201 | } | |
202 | } | |
c04d6cfa AL |
203 | } |
204 | ||
833d4668 AK |
205 | static int spapr_fixup_cpu_smt_dt(void *fdt, int offset, PowerPCCPU *cpu, |
206 | int smt_threads) | |
207 | { | |
208 | int i, ret = 0; | |
209 | uint32_t servers_prop[smt_threads]; | |
210 | uint32_t gservers_prop[smt_threads * 2]; | |
81210c20 | 211 | int index = ppc_get_vcpu_id(cpu); |
833d4668 | 212 | |
d6e166c0 DG |
213 | if (cpu->compat_pvr) { |
214 | ret = fdt_setprop_cell(fdt, offset, "cpu-version", cpu->compat_pvr); | |
6d9412ea AK |
215 | if (ret < 0) { |
216 | return ret; | |
217 | } | |
218 | } | |
219 | ||
833d4668 AK |
220 | /* Build interrupt servers and gservers properties */ |
221 | for (i = 0; i < smt_threads; i++) { | |
222 | servers_prop[i] = cpu_to_be32(index + i); | |
223 | /* Hack, direct the group queues back to cpu 0 */ | |
224 | gservers_prop[i*2] = cpu_to_be32(index + i); | |
225 | gservers_prop[i*2 + 1] = 0; | |
226 | } | |
227 | ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s", | |
228 | servers_prop, sizeof(servers_prop)); | |
229 | if (ret < 0) { | |
230 | return ret; | |
231 | } | |
232 | ret = fdt_setprop(fdt, offset, "ibm,ppc-interrupt-gserver#s", | |
233 | gservers_prop, sizeof(gservers_prop)); | |
234 | ||
235 | return ret; | |
236 | } | |
237 | ||
99861ecb | 238 | static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, PowerPCCPU *cpu) |
0da6f3fe | 239 | { |
81210c20 | 240 | int index = ppc_get_vcpu_id(cpu); |
0da6f3fe BR |
241 | uint32_t associativity[] = {cpu_to_be32(0x5), |
242 | cpu_to_be32(0x0), | |
243 | cpu_to_be32(0x0), | |
244 | cpu_to_be32(0x0), | |
15f8b142 | 245 | cpu_to_be32(cpu->node_id), |
0da6f3fe BR |
246 | cpu_to_be32(index)}; |
247 | ||
248 | /* Advertise NUMA via ibm,associativity */ | |
99861ecb | 249 | return fdt_setprop(fdt, offset, "ibm,associativity", associativity, |
0da6f3fe | 250 | sizeof(associativity)); |
0da6f3fe BR |
251 | } |
252 | ||
86d5771a | 253 | /* Populate the "ibm,pa-features" property */ |
e957f6a9 SB |
254 | static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset, |
255 | bool legacy_guest) | |
86d5771a SB |
256 | { |
257 | uint8_t pa_features_206[] = { 6, 0, | |
258 | 0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 }; | |
259 | uint8_t pa_features_207[] = { 24, 0, | |
260 | 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0, | |
261 | 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, | |
262 | 0x00, 0x00, 0x00, 0x00, 0x80, 0x00, | |
263 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00 }; | |
9fb4541f SB |
264 | uint8_t pa_features_300[] = { 66, 0, |
265 | /* 0: MMU|FPU|SLB|RUN|DABR|NX, 1: fri[nzpm]|DABRX|SPRG3|SLB0|PP110 */ | |
266 | /* 2: VPM|DS205|PPR|DS202|DS206, 3: LSD|URG, SSO, 5: LE|CFAR|EB|LSQ */ | |
267 | 0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0, /* 0 - 5 */ | |
268 | /* 6: DS207 */ | |
269 | 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, /* 6 - 11 */ | |
270 | /* 16: Vector */ | |
86d5771a | 271 | 0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 12 - 17 */ |
9fb4541f | 272 | /* 18: Vec. Scalar, 20: Vec. XOR, 22: HTM */ |
9bf502fe | 273 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 18 - 23 */ |
9fb4541f SB |
274 | /* 24: Ext. Dec, 26: 64 bit ftrs, 28: PM ftrs */ |
275 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 24 - 29 */ | |
276 | /* 30: MMR, 32: LE atomic, 34: EBB + ext EBB */ | |
277 | 0x80, 0x00, 0x80, 0x00, 0xC0, 0x00, /* 30 - 35 */ | |
278 | /* 36: SPR SO, 38: Copy/Paste, 40: Radix MMU */ | |
279 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 36 - 41 */ | |
280 | /* 42: PM, 44: PC RA, 46: SC vec'd */ | |
281 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 42 - 47 */ | |
282 | /* 48: SIMD, 50: QP BFP, 52: String */ | |
283 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 48 - 53 */ | |
284 | /* 54: DecFP, 56: DecI, 58: SHA */ | |
285 | 0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 54 - 59 */ | |
286 | /* 60: NM atomic, 62: RNG */ | |
287 | 0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 60 - 65 */ | |
288 | }; | |
86d5771a SB |
289 | uint8_t *pa_features; |
290 | size_t pa_size; | |
291 | ||
292 | switch (POWERPC_MMU_VER(env->mmu_model)) { | |
293 | case POWERPC_MMU_VER_2_06: | |
294 | pa_features = pa_features_206; | |
295 | pa_size = sizeof(pa_features_206); | |
296 | break; | |
297 | case POWERPC_MMU_VER_2_07: | |
298 | pa_features = pa_features_207; | |
299 | pa_size = sizeof(pa_features_207); | |
300 | break; | |
301 | case POWERPC_MMU_VER_3_00: | |
302 | pa_features = pa_features_300; | |
303 | pa_size = sizeof(pa_features_300); | |
304 | break; | |
305 | default: | |
306 | return; | |
307 | } | |
308 | ||
309 | if (env->ci_large_pages) { | |
310 | /* | |
311 | * Note: we keep CI large pages off by default because a 64K capable | |
312 | * guest provisioned with large pages might otherwise try to map a qemu | |
313 | * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages | |
314 | * even if that qemu runs on a 4k host. | |
315 | * We dd this bit back here if we are confident this is not an issue | |
316 | */ | |
317 | pa_features[3] |= 0x20; | |
318 | } | |
319 | if (kvmppc_has_cap_htm() && pa_size > 24) { | |
320 | pa_features[24] |= 0x80; /* Transactional memory support */ | |
321 | } | |
e957f6a9 SB |
322 | if (legacy_guest && pa_size > 40) { |
323 | /* Workaround for broken kernels that attempt (guest) radix | |
324 | * mode when they can't handle it, if they see the radix bit set | |
325 | * in pa-features. So hide it from them. */ | |
326 | pa_features[40 + 2] &= ~0x80; /* Radix MMU */ | |
327 | } | |
86d5771a SB |
328 | |
329 | _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size))); | |
330 | } | |
331 | ||
28e02042 | 332 | static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr) |
6e806cc3 | 333 | { |
82677ed2 AK |
334 | int ret = 0, offset, cpus_offset; |
335 | CPUState *cs; | |
6e806cc3 BR |
336 | char cpu_model[32]; |
337 | int smt = kvmppc_smt_threads(); | |
7f763a5d | 338 | uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; |
6e806cc3 | 339 | |
82677ed2 AK |
340 | CPU_FOREACH(cs) { |
341 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
e957f6a9 | 342 | CPUPPCState *env = &cpu->env; |
82677ed2 | 343 | DeviceClass *dc = DEVICE_GET_CLASS(cs); |
81210c20 | 344 | int index = ppc_get_vcpu_id(cpu); |
12dbeb16 | 345 | int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); |
6e806cc3 | 346 | |
0f20ba62 | 347 | if ((index % smt) != 0) { |
6e806cc3 BR |
348 | continue; |
349 | } | |
350 | ||
82677ed2 | 351 | snprintf(cpu_model, 32, "%s@%x", dc->fw_name, index); |
6e806cc3 | 352 | |
82677ed2 AK |
353 | cpus_offset = fdt_path_offset(fdt, "/cpus"); |
354 | if (cpus_offset < 0) { | |
355 | cpus_offset = fdt_add_subnode(fdt, fdt_path_offset(fdt, "/"), | |
356 | "cpus"); | |
357 | if (cpus_offset < 0) { | |
358 | return cpus_offset; | |
359 | } | |
360 | } | |
361 | offset = fdt_subnode_offset(fdt, cpus_offset, cpu_model); | |
6e806cc3 | 362 | if (offset < 0) { |
82677ed2 AK |
363 | offset = fdt_add_subnode(fdt, cpus_offset, cpu_model); |
364 | if (offset < 0) { | |
365 | return offset; | |
366 | } | |
6e806cc3 BR |
367 | } |
368 | ||
7f763a5d DG |
369 | ret = fdt_setprop(fdt, offset, "ibm,pft-size", |
370 | pft_size_prop, sizeof(pft_size_prop)); | |
6e806cc3 BR |
371 | if (ret < 0) { |
372 | return ret; | |
373 | } | |
833d4668 | 374 | |
99861ecb IM |
375 | if (nb_numa_nodes > 1) { |
376 | ret = spapr_fixup_cpu_numa_dt(fdt, offset, cpu); | |
377 | if (ret < 0) { | |
378 | return ret; | |
379 | } | |
0da6f3fe BR |
380 | } |
381 | ||
12dbeb16 | 382 | ret = spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt); |
833d4668 AK |
383 | if (ret < 0) { |
384 | return ret; | |
385 | } | |
e957f6a9 SB |
386 | |
387 | spapr_populate_pa_features(env, fdt, offset, | |
388 | spapr->cas_legacy_guest_workaround); | |
6e806cc3 BR |
389 | } |
390 | return ret; | |
391 | } | |
392 | ||
b082d65a AK |
393 | static hwaddr spapr_node0_size(void) |
394 | { | |
fb164994 DG |
395 | MachineState *machine = MACHINE(qdev_get_machine()); |
396 | ||
b082d65a AK |
397 | if (nb_numa_nodes) { |
398 | int i; | |
399 | for (i = 0; i < nb_numa_nodes; ++i) { | |
400 | if (numa_info[i].node_mem) { | |
fb164994 DG |
401 | return MIN(pow2floor(numa_info[i].node_mem), |
402 | machine->ram_size); | |
b082d65a AK |
403 | } |
404 | } | |
405 | } | |
fb164994 | 406 | return machine->ram_size; |
b082d65a AK |
407 | } |
408 | ||
a1d59c0f AK |
409 | static void add_str(GString *s, const gchar *s1) |
410 | { | |
411 | g_string_append_len(s, s1, strlen(s1) + 1); | |
412 | } | |
7f763a5d | 413 | |
03d196b7 | 414 | static int spapr_populate_memory_node(void *fdt, int nodeid, hwaddr start, |
26a8c353 AK |
415 | hwaddr size) |
416 | { | |
417 | uint32_t associativity[] = { | |
418 | cpu_to_be32(0x4), /* length */ | |
419 | cpu_to_be32(0x0), cpu_to_be32(0x0), | |
c3b4f589 | 420 | cpu_to_be32(0x0), cpu_to_be32(nodeid) |
26a8c353 AK |
421 | }; |
422 | char mem_name[32]; | |
423 | uint64_t mem_reg_property[2]; | |
424 | int off; | |
425 | ||
426 | mem_reg_property[0] = cpu_to_be64(start); | |
427 | mem_reg_property[1] = cpu_to_be64(size); | |
428 | ||
429 | sprintf(mem_name, "memory@" TARGET_FMT_lx, start); | |
430 | off = fdt_add_subnode(fdt, 0, mem_name); | |
431 | _FDT(off); | |
432 | _FDT((fdt_setprop_string(fdt, off, "device_type", "memory"))); | |
433 | _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property, | |
434 | sizeof(mem_reg_property)))); | |
435 | _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity, | |
436 | sizeof(associativity)))); | |
03d196b7 | 437 | return off; |
26a8c353 AK |
438 | } |
439 | ||
28e02042 | 440 | static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt) |
7f763a5d | 441 | { |
fb164994 | 442 | MachineState *machine = MACHINE(spapr); |
7db8a127 AK |
443 | hwaddr mem_start, node_size; |
444 | int i, nb_nodes = nb_numa_nodes; | |
445 | NodeInfo *nodes = numa_info; | |
446 | NodeInfo ramnode; | |
447 | ||
448 | /* No NUMA nodes, assume there is just one node with whole RAM */ | |
449 | if (!nb_numa_nodes) { | |
450 | nb_nodes = 1; | |
fb164994 | 451 | ramnode.node_mem = machine->ram_size; |
7db8a127 | 452 | nodes = &ramnode; |
5fe269b1 | 453 | } |
7f763a5d | 454 | |
7db8a127 AK |
455 | for (i = 0, mem_start = 0; i < nb_nodes; ++i) { |
456 | if (!nodes[i].node_mem) { | |
457 | continue; | |
458 | } | |
fb164994 | 459 | if (mem_start >= machine->ram_size) { |
5fe269b1 PM |
460 | node_size = 0; |
461 | } else { | |
7db8a127 | 462 | node_size = nodes[i].node_mem; |
fb164994 DG |
463 | if (node_size > machine->ram_size - mem_start) { |
464 | node_size = machine->ram_size - mem_start; | |
5fe269b1 PM |
465 | } |
466 | } | |
7db8a127 AK |
467 | if (!mem_start) { |
468 | /* ppc_spapr_init() checks for rma_size <= node0_size already */ | |
e8f986fc | 469 | spapr_populate_memory_node(fdt, i, 0, spapr->rma_size); |
7db8a127 AK |
470 | mem_start += spapr->rma_size; |
471 | node_size -= spapr->rma_size; | |
472 | } | |
6010818c AK |
473 | for ( ; node_size; ) { |
474 | hwaddr sizetmp = pow2floor(node_size); | |
475 | ||
476 | /* mem_start != 0 here */ | |
477 | if (ctzl(mem_start) < ctzl(sizetmp)) { | |
478 | sizetmp = 1ULL << ctzl(mem_start); | |
479 | } | |
480 | ||
481 | spapr_populate_memory_node(fdt, i, mem_start, sizetmp); | |
482 | node_size -= sizetmp; | |
483 | mem_start += sizetmp; | |
484 | } | |
7f763a5d DG |
485 | } |
486 | ||
487 | return 0; | |
488 | } | |
489 | ||
0da6f3fe BR |
490 | static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset, |
491 | sPAPRMachineState *spapr) | |
492 | { | |
493 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
494 | CPUPPCState *env = &cpu->env; | |
495 | PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs); | |
81210c20 | 496 | int index = ppc_get_vcpu_id(cpu); |
0da6f3fe BR |
497 | uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40), |
498 | 0xffffffff, 0xffffffff}; | |
afd10a0f BR |
499 | uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() |
500 | : SPAPR_TIMEBASE_FREQ; | |
0da6f3fe BR |
501 | uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000; |
502 | uint32_t page_sizes_prop[64]; | |
503 | size_t page_sizes_prop_size; | |
22419c2a | 504 | uint32_t vcpus_per_socket = smp_threads * smp_cores; |
0da6f3fe | 505 | uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)}; |
12dbeb16 | 506 | int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu)); |
af81cf32 | 507 | sPAPRDRConnector *drc; |
af81cf32 | 508 | int drc_index; |
c64abd1f SB |
509 | uint32_t radix_AP_encodings[PPC_PAGE_SIZES_MAX_SZ]; |
510 | int i; | |
af81cf32 | 511 | |
fbf55397 | 512 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index); |
af81cf32 | 513 | if (drc) { |
0b55aa91 | 514 | drc_index = spapr_drc_index(drc); |
af81cf32 BR |
515 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index))); |
516 | } | |
0da6f3fe BR |
517 | |
518 | _FDT((fdt_setprop_cell(fdt, offset, "reg", index))); | |
519 | _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu"))); | |
520 | ||
521 | _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR]))); | |
522 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size", | |
523 | env->dcache_line_size))); | |
524 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size", | |
525 | env->dcache_line_size))); | |
526 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size", | |
527 | env->icache_line_size))); | |
528 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size", | |
529 | env->icache_line_size))); | |
530 | ||
531 | if (pcc->l1_dcache_size) { | |
532 | _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size", | |
533 | pcc->l1_dcache_size))); | |
534 | } else { | |
3dc6f869 | 535 | warn_report("Unknown L1 dcache size for cpu"); |
0da6f3fe BR |
536 | } |
537 | if (pcc->l1_icache_size) { | |
538 | _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size", | |
539 | pcc->l1_icache_size))); | |
540 | } else { | |
3dc6f869 | 541 | warn_report("Unknown L1 icache size for cpu"); |
0da6f3fe BR |
542 | } |
543 | ||
544 | _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq))); | |
545 | _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq))); | |
fd5da5c4 | 546 | _FDT((fdt_setprop_cell(fdt, offset, "slb-size", env->slb_nr))); |
0da6f3fe BR |
547 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", env->slb_nr))); |
548 | _FDT((fdt_setprop_string(fdt, offset, "status", "okay"))); | |
549 | _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0))); | |
550 | ||
551 | if (env->spr_cb[SPR_PURR].oea_read) { | |
552 | _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0))); | |
553 | } | |
554 | ||
555 | if (env->mmu_model & POWERPC_MMU_1TSEG) { | |
556 | _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes", | |
557 | segs, sizeof(segs)))); | |
558 | } | |
559 | ||
560 | /* Advertise VMX/VSX (vector extensions) if available | |
561 | * 0 / no property == no vector extensions | |
562 | * 1 == VMX / Altivec available | |
563 | * 2 == VSX available */ | |
564 | if (env->insns_flags & PPC_ALTIVEC) { | |
565 | uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1; | |
566 | ||
567 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx))); | |
568 | } | |
569 | ||
570 | /* Advertise DFP (Decimal Floating Point) if available | |
571 | * 0 / no property == no DFP | |
572 | * 1 == DFP available */ | |
573 | if (env->insns_flags2 & PPC2_DFP) { | |
574 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1))); | |
575 | } | |
576 | ||
3654fa95 | 577 | page_sizes_prop_size = ppc_create_page_sizes_prop(env, page_sizes_prop, |
0da6f3fe BR |
578 | sizeof(page_sizes_prop)); |
579 | if (page_sizes_prop_size) { | |
580 | _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes", | |
581 | page_sizes_prop, page_sizes_prop_size))); | |
582 | } | |
583 | ||
e957f6a9 | 584 | spapr_populate_pa_features(env, fdt, offset, false); |
90da0d5a | 585 | |
0da6f3fe | 586 | _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", |
22419c2a | 587 | cs->cpu_index / vcpus_per_socket))); |
0da6f3fe BR |
588 | |
589 | _FDT((fdt_setprop(fdt, offset, "ibm,pft-size", | |
590 | pft_size_prop, sizeof(pft_size_prop)))); | |
591 | ||
99861ecb IM |
592 | if (nb_numa_nodes > 1) { |
593 | _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cpu)); | |
594 | } | |
0da6f3fe | 595 | |
12dbeb16 | 596 | _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt)); |
c64abd1f SB |
597 | |
598 | if (pcc->radix_page_info) { | |
599 | for (i = 0; i < pcc->radix_page_info->count; i++) { | |
600 | radix_AP_encodings[i] = | |
601 | cpu_to_be32(pcc->radix_page_info->entries[i]); | |
602 | } | |
603 | _FDT((fdt_setprop(fdt, offset, "ibm,processor-radix-AP-encodings", | |
604 | radix_AP_encodings, | |
605 | pcc->radix_page_info->count * | |
606 | sizeof(radix_AP_encodings[0])))); | |
607 | } | |
0da6f3fe BR |
608 | } |
609 | ||
610 | static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr) | |
611 | { | |
612 | CPUState *cs; | |
613 | int cpus_offset; | |
614 | char *nodename; | |
615 | int smt = kvmppc_smt_threads(); | |
616 | ||
617 | cpus_offset = fdt_add_subnode(fdt, 0, "cpus"); | |
618 | _FDT(cpus_offset); | |
619 | _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1))); | |
620 | _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0))); | |
621 | ||
622 | /* | |
623 | * We walk the CPUs in reverse order to ensure that CPU DT nodes | |
624 | * created by fdt_add_subnode() end up in the right order in FDT | |
625 | * for the guest kernel the enumerate the CPUs correctly. | |
626 | */ | |
627 | CPU_FOREACH_REVERSE(cs) { | |
628 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
81210c20 | 629 | int index = ppc_get_vcpu_id(cpu); |
0da6f3fe BR |
630 | DeviceClass *dc = DEVICE_GET_CLASS(cs); |
631 | int offset; | |
632 | ||
633 | if ((index % smt) != 0) { | |
634 | continue; | |
635 | } | |
636 | ||
637 | nodename = g_strdup_printf("%s@%x", dc->fw_name, index); | |
638 | offset = fdt_add_subnode(fdt, cpus_offset, nodename); | |
639 | g_free(nodename); | |
640 | _FDT(offset); | |
641 | spapr_populate_cpu_dt(cs, fdt, offset, spapr); | |
642 | } | |
643 | ||
644 | } | |
645 | ||
03d196b7 BR |
646 | /* |
647 | * Adds ibm,dynamic-reconfiguration-memory node. | |
648 | * Refer to docs/specs/ppc-spapr-hotplug.txt for the documentation | |
649 | * of this device tree node. | |
650 | */ | |
651 | static int spapr_populate_drconf_memory(sPAPRMachineState *spapr, void *fdt) | |
652 | { | |
653 | MachineState *machine = MACHINE(spapr); | |
654 | int ret, i, offset; | |
655 | uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; | |
656 | uint32_t prop_lmb_size[] = {0, cpu_to_be32(lmb_size)}; | |
d0e5a8f2 BR |
657 | uint32_t hotplug_lmb_start = spapr->hotplug_memory.base / lmb_size; |
658 | uint32_t nr_lmbs = (spapr->hotplug_memory.base + | |
659 | memory_region_size(&spapr->hotplug_memory.mr)) / | |
660 | lmb_size; | |
03d196b7 | 661 | uint32_t *int_buf, *cur_index, buf_len; |
6663864e | 662 | int nr_nodes = nb_numa_nodes ? nb_numa_nodes : 1; |
03d196b7 | 663 | |
16c25aef | 664 | /* |
d0e5a8f2 | 665 | * Don't create the node if there is no hotpluggable memory |
16c25aef | 666 | */ |
d0e5a8f2 | 667 | if (machine->ram_size == machine->maxram_size) { |
16c25aef BR |
668 | return 0; |
669 | } | |
670 | ||
ef001f06 TH |
671 | /* |
672 | * Allocate enough buffer size to fit in ibm,dynamic-memory | |
673 | * or ibm,associativity-lookup-arrays | |
674 | */ | |
675 | buf_len = MAX(nr_lmbs * SPAPR_DR_LMB_LIST_ENTRY_SIZE + 1, nr_nodes * 4 + 2) | |
676 | * sizeof(uint32_t); | |
03d196b7 BR |
677 | cur_index = int_buf = g_malloc0(buf_len); |
678 | ||
679 | offset = fdt_add_subnode(fdt, 0, "ibm,dynamic-reconfiguration-memory"); | |
680 | ||
681 | ret = fdt_setprop(fdt, offset, "ibm,lmb-size", prop_lmb_size, | |
682 | sizeof(prop_lmb_size)); | |
683 | if (ret < 0) { | |
684 | goto out; | |
685 | } | |
686 | ||
687 | ret = fdt_setprop_cell(fdt, offset, "ibm,memory-flags-mask", 0xff); | |
688 | if (ret < 0) { | |
689 | goto out; | |
690 | } | |
691 | ||
692 | ret = fdt_setprop_cell(fdt, offset, "ibm,memory-preservation-time", 0x0); | |
693 | if (ret < 0) { | |
694 | goto out; | |
695 | } | |
696 | ||
697 | /* ibm,dynamic-memory */ | |
698 | int_buf[0] = cpu_to_be32(nr_lmbs); | |
699 | cur_index++; | |
700 | for (i = 0; i < nr_lmbs; i++) { | |
d0e5a8f2 | 701 | uint64_t addr = i * lmb_size; |
03d196b7 BR |
702 | uint32_t *dynamic_memory = cur_index; |
703 | ||
d0e5a8f2 BR |
704 | if (i >= hotplug_lmb_start) { |
705 | sPAPRDRConnector *drc; | |
d0e5a8f2 | 706 | |
fbf55397 | 707 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, i); |
d0e5a8f2 | 708 | g_assert(drc); |
d0e5a8f2 BR |
709 | |
710 | dynamic_memory[0] = cpu_to_be32(addr >> 32); | |
711 | dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); | |
0b55aa91 | 712 | dynamic_memory[2] = cpu_to_be32(spapr_drc_index(drc)); |
d0e5a8f2 BR |
713 | dynamic_memory[3] = cpu_to_be32(0); /* reserved */ |
714 | dynamic_memory[4] = cpu_to_be32(numa_get_node(addr, NULL)); | |
715 | if (memory_region_present(get_system_memory(), addr)) { | |
716 | dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_ASSIGNED); | |
717 | } else { | |
718 | dynamic_memory[5] = cpu_to_be32(0); | |
719 | } | |
03d196b7 | 720 | } else { |
d0e5a8f2 BR |
721 | /* |
722 | * LMB information for RMA, boot time RAM and gap b/n RAM and | |
723 | * hotplug memory region -- all these are marked as reserved | |
724 | * and as having no valid DRC. | |
725 | */ | |
726 | dynamic_memory[0] = cpu_to_be32(addr >> 32); | |
727 | dynamic_memory[1] = cpu_to_be32(addr & 0xffffffff); | |
728 | dynamic_memory[2] = cpu_to_be32(0); | |
729 | dynamic_memory[3] = cpu_to_be32(0); /* reserved */ | |
730 | dynamic_memory[4] = cpu_to_be32(-1); | |
731 | dynamic_memory[5] = cpu_to_be32(SPAPR_LMB_FLAGS_RESERVED | | |
732 | SPAPR_LMB_FLAGS_DRC_INVALID); | |
03d196b7 BR |
733 | } |
734 | ||
735 | cur_index += SPAPR_DR_LMB_LIST_ENTRY_SIZE; | |
736 | } | |
737 | ret = fdt_setprop(fdt, offset, "ibm,dynamic-memory", int_buf, buf_len); | |
738 | if (ret < 0) { | |
739 | goto out; | |
740 | } | |
741 | ||
742 | /* ibm,associativity-lookup-arrays */ | |
743 | cur_index = int_buf; | |
6663864e | 744 | int_buf[0] = cpu_to_be32(nr_nodes); |
03d196b7 BR |
745 | int_buf[1] = cpu_to_be32(4); /* Number of entries per associativity list */ |
746 | cur_index += 2; | |
6663864e | 747 | for (i = 0; i < nr_nodes; i++) { |
03d196b7 BR |
748 | uint32_t associativity[] = { |
749 | cpu_to_be32(0x0), | |
750 | cpu_to_be32(0x0), | |
751 | cpu_to_be32(0x0), | |
752 | cpu_to_be32(i) | |
753 | }; | |
754 | memcpy(cur_index, associativity, sizeof(associativity)); | |
755 | cur_index += 4; | |
756 | } | |
757 | ret = fdt_setprop(fdt, offset, "ibm,associativity-lookup-arrays", int_buf, | |
758 | (cur_index - int_buf) * sizeof(uint32_t)); | |
759 | out: | |
760 | g_free(int_buf); | |
761 | return ret; | |
762 | } | |
763 | ||
6787d27b MR |
764 | static int spapr_dt_cas_updates(sPAPRMachineState *spapr, void *fdt, |
765 | sPAPROptionVector *ov5_updates) | |
766 | { | |
767 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr); | |
417ece33 | 768 | int ret = 0, offset; |
6787d27b MR |
769 | |
770 | /* Generate ibm,dynamic-reconfiguration-memory node if required */ | |
771 | if (spapr_ovec_test(ov5_updates, OV5_DRCONF_MEMORY)) { | |
772 | g_assert(smc->dr_lmb_enabled); | |
773 | ret = spapr_populate_drconf_memory(spapr, fdt); | |
417ece33 MR |
774 | if (ret) { |
775 | goto out; | |
776 | } | |
6787d27b MR |
777 | } |
778 | ||
417ece33 MR |
779 | offset = fdt_path_offset(fdt, "/chosen"); |
780 | if (offset < 0) { | |
781 | offset = fdt_add_subnode(fdt, 0, "chosen"); | |
782 | if (offset < 0) { | |
783 | return offset; | |
784 | } | |
785 | } | |
786 | ret = spapr_ovec_populate_dt(fdt, offset, spapr->ov5_cas, | |
787 | "ibm,architecture-vec-5"); | |
788 | ||
789 | out: | |
6787d27b MR |
790 | return ret; |
791 | } | |
792 | ||
10f12e64 DHB |
793 | static bool spapr_hotplugged_dev_before_cas(void) |
794 | { | |
795 | Object *drc_container, *obj; | |
796 | ObjectProperty *prop; | |
797 | ObjectPropertyIterator iter; | |
798 | ||
799 | drc_container = container_get(object_get_root(), "/dr-connector"); | |
800 | object_property_iter_init(&iter, drc_container); | |
801 | while ((prop = object_property_iter_next(&iter))) { | |
802 | if (!strstart(prop->type, "link<", NULL)) { | |
803 | continue; | |
804 | } | |
805 | obj = object_property_get_link(drc_container, prop->name, NULL); | |
806 | if (spapr_drc_needed(obj)) { | |
807 | return true; | |
808 | } | |
809 | } | |
810 | return false; | |
811 | } | |
812 | ||
03d196b7 BR |
813 | int spapr_h_cas_compose_response(sPAPRMachineState *spapr, |
814 | target_ulong addr, target_ulong size, | |
6787d27b | 815 | sPAPROptionVector *ov5_updates) |
03d196b7 BR |
816 | { |
817 | void *fdt, *fdt_skel; | |
818 | sPAPRDeviceTreeUpdateHeader hdr = { .version_id = 1 }; | |
03d196b7 | 819 | |
10f12e64 DHB |
820 | if (spapr_hotplugged_dev_before_cas()) { |
821 | return 1; | |
822 | } | |
823 | ||
03d196b7 BR |
824 | size -= sizeof(hdr); |
825 | ||
10f12e64 | 826 | /* Create skeleton */ |
03d196b7 BR |
827 | fdt_skel = g_malloc0(size); |
828 | _FDT((fdt_create(fdt_skel, size))); | |
829 | _FDT((fdt_begin_node(fdt_skel, ""))); | |
830 | _FDT((fdt_end_node(fdt_skel))); | |
831 | _FDT((fdt_finish(fdt_skel))); | |
832 | fdt = g_malloc0(size); | |
833 | _FDT((fdt_open_into(fdt_skel, fdt, size))); | |
834 | g_free(fdt_skel); | |
835 | ||
836 | /* Fixup cpu nodes */ | |
5b120785 | 837 | _FDT((spapr_fixup_cpu_dt(fdt, spapr))); |
03d196b7 | 838 | |
6787d27b MR |
839 | if (spapr_dt_cas_updates(spapr, fdt, ov5_updates)) { |
840 | return -1; | |
03d196b7 BR |
841 | } |
842 | ||
843 | /* Pack resulting tree */ | |
844 | _FDT((fdt_pack(fdt))); | |
845 | ||
846 | if (fdt_totalsize(fdt) + sizeof(hdr) > size) { | |
847 | trace_spapr_cas_failed(size); | |
848 | return -1; | |
849 | } | |
850 | ||
851 | cpu_physical_memory_write(addr, &hdr, sizeof(hdr)); | |
852 | cpu_physical_memory_write(addr + sizeof(hdr), fdt, fdt_totalsize(fdt)); | |
853 | trace_spapr_cas_continue(fdt_totalsize(fdt) + sizeof(hdr)); | |
854 | g_free(fdt); | |
855 | ||
856 | return 0; | |
857 | } | |
858 | ||
3f5dabce DG |
859 | static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt) |
860 | { | |
861 | int rtas; | |
862 | GString *hypertas = g_string_sized_new(256); | |
863 | GString *qemu_hypertas = g_string_sized_new(256); | |
864 | uint32_t refpoints[] = { cpu_to_be32(0x4), cpu_to_be32(0x4) }; | |
865 | uint64_t max_hotplug_addr = spapr->hotplug_memory.base + | |
866 | memory_region_size(&spapr->hotplug_memory.mr); | |
867 | uint32_t lrdr_capacity[] = { | |
868 | cpu_to_be32(max_hotplug_addr >> 32), | |
869 | cpu_to_be32(max_hotplug_addr & 0xffffffff), | |
870 | 0, cpu_to_be32(SPAPR_MEMORY_BLOCK_SIZE), | |
871 | cpu_to_be32(max_cpus / smp_threads), | |
872 | }; | |
873 | ||
874 | _FDT(rtas = fdt_add_subnode(fdt, 0, "rtas")); | |
875 | ||
876 | /* hypertas */ | |
877 | add_str(hypertas, "hcall-pft"); | |
878 | add_str(hypertas, "hcall-term"); | |
879 | add_str(hypertas, "hcall-dabr"); | |
880 | add_str(hypertas, "hcall-interrupt"); | |
881 | add_str(hypertas, "hcall-tce"); | |
882 | add_str(hypertas, "hcall-vio"); | |
883 | add_str(hypertas, "hcall-splpar"); | |
884 | add_str(hypertas, "hcall-bulk"); | |
885 | add_str(hypertas, "hcall-set-mode"); | |
886 | add_str(hypertas, "hcall-sprg0"); | |
887 | add_str(hypertas, "hcall-copy"); | |
888 | add_str(hypertas, "hcall-debug"); | |
889 | add_str(qemu_hypertas, "hcall-memop1"); | |
890 | ||
891 | if (!kvm_enabled() || kvmppc_spapr_use_multitce()) { | |
892 | add_str(hypertas, "hcall-multi-tce"); | |
893 | } | |
30f4b05b DG |
894 | |
895 | if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) { | |
896 | add_str(hypertas, "hcall-hpt-resize"); | |
897 | } | |
898 | ||
3f5dabce DG |
899 | _FDT(fdt_setprop(fdt, rtas, "ibm,hypertas-functions", |
900 | hypertas->str, hypertas->len)); | |
901 | g_string_free(hypertas, TRUE); | |
902 | _FDT(fdt_setprop(fdt, rtas, "qemu,hypertas-functions", | |
903 | qemu_hypertas->str, qemu_hypertas->len)); | |
904 | g_string_free(qemu_hypertas, TRUE); | |
905 | ||
906 | _FDT(fdt_setprop(fdt, rtas, "ibm,associativity-reference-points", | |
907 | refpoints, sizeof(refpoints))); | |
908 | ||
909 | _FDT(fdt_setprop_cell(fdt, rtas, "rtas-error-log-max", | |
910 | RTAS_ERROR_LOG_MAX)); | |
911 | _FDT(fdt_setprop_cell(fdt, rtas, "rtas-event-scan-rate", | |
912 | RTAS_EVENT_SCAN_RATE)); | |
913 | ||
914 | if (msi_nonbroken) { | |
915 | _FDT(fdt_setprop(fdt, rtas, "ibm,change-msix-capable", NULL, 0)); | |
916 | } | |
917 | ||
918 | /* | |
919 | * According to PAPR, rtas ibm,os-term does not guarantee a return | |
920 | * back to the guest cpu. | |
921 | * | |
922 | * While an additional ibm,extended-os-term property indicates | |
923 | * that rtas call return will always occur. Set this property. | |
924 | */ | |
925 | _FDT(fdt_setprop(fdt, rtas, "ibm,extended-os-term", NULL, 0)); | |
926 | ||
927 | _FDT(fdt_setprop(fdt, rtas, "ibm,lrdr-capacity", | |
928 | lrdr_capacity, sizeof(lrdr_capacity))); | |
929 | ||
930 | spapr_dt_rtas_tokens(fdt, rtas); | |
931 | } | |
932 | ||
9fb4541f SB |
933 | /* Prepare ibm,arch-vec-5-platform-support, which indicates the MMU features |
934 | * that the guest may request and thus the valid values for bytes 24..26 of | |
935 | * option vector 5: */ | |
936 | static void spapr_dt_ov5_platform_support(void *fdt, int chosen) | |
937 | { | |
545d6e2b SJS |
938 | PowerPCCPU *first_ppc_cpu = POWERPC_CPU(first_cpu); |
939 | ||
f2b14e3a CLG |
940 | char val[2 * 4] = { |
941 | 23, 0x00, /* Xive mode: 0 = legacy (as in ISA 2.7), 1 = Exploitation */ | |
9fb4541f SB |
942 | 24, 0x00, /* Hash/Radix, filled in below. */ |
943 | 25, 0x00, /* Hash options: Segment Tables == no, GTSE == no. */ | |
944 | 26, 0x40, /* Radix options: GTSE == yes. */ | |
945 | }; | |
946 | ||
947 | if (kvm_enabled()) { | |
948 | if (kvmppc_has_cap_mmu_radix() && kvmppc_has_cap_mmu_hash_v3()) { | |
f2b14e3a | 949 | val[3] = 0x80; /* OV5_MMU_BOTH */ |
9fb4541f | 950 | } else if (kvmppc_has_cap_mmu_radix()) { |
f2b14e3a | 951 | val[3] = 0x40; /* OV5_MMU_RADIX_300 */ |
9fb4541f | 952 | } else { |
f2b14e3a | 953 | val[3] = 0x00; /* Hash */ |
9fb4541f SB |
954 | } |
955 | } else { | |
545d6e2b SJS |
956 | if (first_ppc_cpu->env.mmu_model & POWERPC_MMU_V3) { |
957 | /* V3 MMU supports both hash and radix (with dynamic switching) */ | |
f2b14e3a | 958 | val[3] = 0xC0; |
545d6e2b SJS |
959 | } else { |
960 | /* Otherwise we can only do hash */ | |
f2b14e3a | 961 | val[3] = 0x00; |
545d6e2b | 962 | } |
9fb4541f SB |
963 | } |
964 | _FDT(fdt_setprop(fdt, chosen, "ibm,arch-vec-5-platform-support", | |
965 | val, sizeof(val))); | |
966 | } | |
967 | ||
7c866c6a DG |
968 | static void spapr_dt_chosen(sPAPRMachineState *spapr, void *fdt) |
969 | { | |
970 | MachineState *machine = MACHINE(spapr); | |
971 | int chosen; | |
972 | const char *boot_device = machine->boot_order; | |
973 | char *stdout_path = spapr_vio_stdout_path(spapr->vio_bus); | |
974 | size_t cb = 0; | |
975 | char *bootlist = get_boot_devices_list(&cb, true); | |
7c866c6a DG |
976 | |
977 | _FDT(chosen = fdt_add_subnode(fdt, 0, "chosen")); | |
978 | ||
7c866c6a DG |
979 | _FDT(fdt_setprop_string(fdt, chosen, "bootargs", machine->kernel_cmdline)); |
980 | _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-start", | |
981 | spapr->initrd_base)); | |
982 | _FDT(fdt_setprop_cell(fdt, chosen, "linux,initrd-end", | |
983 | spapr->initrd_base + spapr->initrd_size)); | |
984 | ||
985 | if (spapr->kernel_size) { | |
986 | uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR), | |
987 | cpu_to_be64(spapr->kernel_size) }; | |
988 | ||
989 | _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel", | |
990 | &kprop, sizeof(kprop))); | |
991 | if (spapr->kernel_le) { | |
992 | _FDT(fdt_setprop(fdt, chosen, "qemu,boot-kernel-le", NULL, 0)); | |
993 | } | |
994 | } | |
995 | if (boot_menu) { | |
996 | _FDT((fdt_setprop_cell(fdt, chosen, "qemu,boot-menu", boot_menu))); | |
997 | } | |
998 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-width", graphic_width)); | |
999 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-height", graphic_height)); | |
1000 | _FDT(fdt_setprop_cell(fdt, chosen, "qemu,graphic-depth", graphic_depth)); | |
1001 | ||
1002 | if (cb && bootlist) { | |
1003 | int i; | |
1004 | ||
1005 | for (i = 0; i < cb; i++) { | |
1006 | if (bootlist[i] == '\n') { | |
1007 | bootlist[i] = ' '; | |
1008 | } | |
1009 | } | |
1010 | _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-list", bootlist)); | |
1011 | } | |
1012 | ||
1013 | if (boot_device && strlen(boot_device)) { | |
1014 | _FDT(fdt_setprop_string(fdt, chosen, "qemu,boot-device", boot_device)); | |
1015 | } | |
1016 | ||
1017 | if (!spapr->has_graphics && stdout_path) { | |
1018 | _FDT(fdt_setprop_string(fdt, chosen, "linux,stdout-path", stdout_path)); | |
1019 | } | |
1020 | ||
9fb4541f SB |
1021 | spapr_dt_ov5_platform_support(fdt, chosen); |
1022 | ||
7c866c6a DG |
1023 | g_free(stdout_path); |
1024 | g_free(bootlist); | |
1025 | } | |
1026 | ||
fca5f2dc DG |
1027 | static void spapr_dt_hypervisor(sPAPRMachineState *spapr, void *fdt) |
1028 | { | |
1029 | /* The /hypervisor node isn't in PAPR - this is a hack to allow PR | |
1030 | * KVM to work under pHyp with some guest co-operation */ | |
1031 | int hypervisor; | |
1032 | uint8_t hypercall[16]; | |
1033 | ||
1034 | _FDT(hypervisor = fdt_add_subnode(fdt, 0, "hypervisor")); | |
1035 | /* indicate KVM hypercall interface */ | |
1036 | _FDT(fdt_setprop_string(fdt, hypervisor, "compatible", "linux,kvm")); | |
1037 | if (kvmppc_has_cap_fixup_hcalls()) { | |
1038 | /* | |
1039 | * Older KVM versions with older guest kernels were broken | |
1040 | * with the magic page, don't allow the guest to map it. | |
1041 | */ | |
1042 | if (!kvmppc_get_hypercall(first_cpu->env_ptr, hypercall, | |
1043 | sizeof(hypercall))) { | |
1044 | _FDT(fdt_setprop(fdt, hypervisor, "hcall-instructions", | |
1045 | hypercall, sizeof(hypercall))); | |
1046 | } | |
1047 | } | |
1048 | } | |
1049 | ||
997b6cfc DG |
1050 | static void *spapr_build_fdt(sPAPRMachineState *spapr, |
1051 | hwaddr rtas_addr, | |
1052 | hwaddr rtas_size) | |
a3467baa | 1053 | { |
5b2128d2 | 1054 | MachineState *machine = MACHINE(qdev_get_machine()); |
3c0c47e3 | 1055 | MachineClass *mc = MACHINE_GET_CLASS(machine); |
c20d332a | 1056 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); |
7c866c6a | 1057 | int ret; |
a3467baa | 1058 | void *fdt; |
3384f95c | 1059 | sPAPRPHBState *phb; |
398a0bd5 | 1060 | char *buf; |
a3467baa | 1061 | |
398a0bd5 DG |
1062 | fdt = g_malloc0(FDT_MAX_SIZE); |
1063 | _FDT((fdt_create_empty_tree(fdt, FDT_MAX_SIZE))); | |
a3467baa | 1064 | |
398a0bd5 DG |
1065 | /* Root node */ |
1066 | _FDT(fdt_setprop_string(fdt, 0, "device_type", "chrp")); | |
1067 | _FDT(fdt_setprop_string(fdt, 0, "model", "IBM pSeries (emulated by qemu)")); | |
1068 | _FDT(fdt_setprop_string(fdt, 0, "compatible", "qemu,pseries")); | |
1069 | ||
1070 | /* | |
1071 | * Add info to guest to indentify which host is it being run on | |
1072 | * and what is the uuid of the guest | |
1073 | */ | |
1074 | if (kvmppc_get_host_model(&buf)) { | |
1075 | _FDT(fdt_setprop_string(fdt, 0, "host-model", buf)); | |
1076 | g_free(buf); | |
1077 | } | |
1078 | if (kvmppc_get_host_serial(&buf)) { | |
1079 | _FDT(fdt_setprop_string(fdt, 0, "host-serial", buf)); | |
1080 | g_free(buf); | |
1081 | } | |
1082 | ||
1083 | buf = qemu_uuid_unparse_strdup(&qemu_uuid); | |
1084 | ||
1085 | _FDT(fdt_setprop_string(fdt, 0, "vm,uuid", buf)); | |
1086 | if (qemu_uuid_set) { | |
1087 | _FDT(fdt_setprop_string(fdt, 0, "system-id", buf)); | |
1088 | } | |
1089 | g_free(buf); | |
1090 | ||
1091 | if (qemu_get_vm_name()) { | |
1092 | _FDT(fdt_setprop_string(fdt, 0, "ibm,partition-name", | |
1093 | qemu_get_vm_name())); | |
1094 | } | |
1095 | ||
1096 | _FDT(fdt_setprop_cell(fdt, 0, "#address-cells", 2)); | |
1097 | _FDT(fdt_setprop_cell(fdt, 0, "#size-cells", 2)); | |
4040ab72 | 1098 | |
fc7e0765 DG |
1099 | /* /interrupt controller */ |
1100 | spapr_dt_xics(xics_max_server_number(), fdt, PHANDLE_XICP); | |
1101 | ||
e8f986fc BR |
1102 | ret = spapr_populate_memory(spapr, fdt); |
1103 | if (ret < 0) { | |
ce9863b7 | 1104 | error_report("couldn't setup memory nodes in fdt"); |
e8f986fc | 1105 | exit(1); |
7f763a5d DG |
1106 | } |
1107 | ||
bf5a6696 DG |
1108 | /* /vdevice */ |
1109 | spapr_dt_vdevice(spapr->vio_bus, fdt); | |
4040ab72 | 1110 | |
4d9392be TH |
1111 | if (object_resolve_path_type("", TYPE_SPAPR_RNG, NULL)) { |
1112 | ret = spapr_rng_populate_dt(fdt); | |
1113 | if (ret < 0) { | |
ce9863b7 | 1114 | error_report("could not set up rng device in the fdt"); |
4d9392be TH |
1115 | exit(1); |
1116 | } | |
1117 | } | |
1118 | ||
3384f95c | 1119 | QLIST_FOREACH(phb, &spapr->phbs, list) { |
e0fdbd7c | 1120 | ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt); |
da34fed7 TH |
1121 | if (ret < 0) { |
1122 | error_report("couldn't setup PCI devices in fdt"); | |
1123 | exit(1); | |
1124 | } | |
3384f95c DG |
1125 | } |
1126 | ||
0da6f3fe BR |
1127 | /* cpus */ |
1128 | spapr_populate_cpus_dt_node(fdt, spapr); | |
6e806cc3 | 1129 | |
c20d332a BR |
1130 | if (smc->dr_lmb_enabled) { |
1131 | _FDT(spapr_drc_populate_dt(fdt, 0, NULL, SPAPR_DR_CONNECTOR_TYPE_LMB)); | |
1132 | } | |
1133 | ||
c5514d0e | 1134 | if (mc->has_hotpluggable_cpus) { |
af81cf32 BR |
1135 | int offset = fdt_path_offset(fdt, "/cpus"); |
1136 | ret = spapr_drc_populate_dt(fdt, offset, NULL, | |
1137 | SPAPR_DR_CONNECTOR_TYPE_CPU); | |
1138 | if (ret < 0) { | |
1139 | error_report("Couldn't set up CPU DR device tree properties"); | |
1140 | exit(1); | |
1141 | } | |
1142 | } | |
1143 | ||
ffb1e275 | 1144 | /* /event-sources */ |
ffbb1705 | 1145 | spapr_dt_events(spapr, fdt); |
ffb1e275 | 1146 | |
3f5dabce DG |
1147 | /* /rtas */ |
1148 | spapr_dt_rtas(spapr, fdt); | |
1149 | ||
7c866c6a DG |
1150 | /* /chosen */ |
1151 | spapr_dt_chosen(spapr, fdt); | |
cf6e5223 | 1152 | |
fca5f2dc DG |
1153 | /* /hypervisor */ |
1154 | if (kvm_enabled()) { | |
1155 | spapr_dt_hypervisor(spapr, fdt); | |
1156 | } | |
1157 | ||
cf6e5223 DG |
1158 | /* Build memory reserve map */ |
1159 | if (spapr->kernel_size) { | |
1160 | _FDT((fdt_add_mem_rsv(fdt, KERNEL_LOAD_ADDR, spapr->kernel_size))); | |
1161 | } | |
1162 | if (spapr->initrd_size) { | |
1163 | _FDT((fdt_add_mem_rsv(fdt, spapr->initrd_base, spapr->initrd_size))); | |
1164 | } | |
1165 | ||
6787d27b MR |
1166 | /* ibm,client-architecture-support updates */ |
1167 | ret = spapr_dt_cas_updates(spapr, fdt, spapr->ov5_cas); | |
1168 | if (ret < 0) { | |
1169 | error_report("couldn't setup CAS properties fdt"); | |
1170 | exit(1); | |
1171 | } | |
1172 | ||
997b6cfc | 1173 | return fdt; |
9fdf0c29 DG |
1174 | } |
1175 | ||
1176 | static uint64_t translate_kernel_address(void *opaque, uint64_t addr) | |
1177 | { | |
1178 | return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR; | |
1179 | } | |
1180 | ||
1d1be34d DG |
1181 | static void emulate_spapr_hypercall(PPCVirtualHypervisor *vhyp, |
1182 | PowerPCCPU *cpu) | |
9fdf0c29 | 1183 | { |
1b14670a AF |
1184 | CPUPPCState *env = &cpu->env; |
1185 | ||
8d04fb55 JK |
1186 | /* The TCG path should also be holding the BQL at this point */ |
1187 | g_assert(qemu_mutex_iothread_locked()); | |
1188 | ||
efcb9383 DG |
1189 | if (msr_pr) { |
1190 | hcall_dprintf("Hypercall made with MSR[PR]=1\n"); | |
1191 | env->gpr[3] = H_PRIVILEGE; | |
1192 | } else { | |
aa100fa4 | 1193 | env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]); |
efcb9383 | 1194 | } |
9fdf0c29 DG |
1195 | } |
1196 | ||
9861bb3e SJS |
1197 | static uint64_t spapr_get_patbe(PPCVirtualHypervisor *vhyp) |
1198 | { | |
1199 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1200 | ||
1201 | return spapr->patb_entry; | |
1202 | } | |
1203 | ||
e6b8fd24 SMJ |
1204 | #define HPTE(_table, _i) (void *)(((uint64_t *)(_table)) + ((_i) * 2)) |
1205 | #define HPTE_VALID(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_VALID) | |
1206 | #define HPTE_DIRTY(_hpte) (tswap64(*((uint64_t *)(_hpte))) & HPTE64_V_HPTE_DIRTY) | |
1207 | #define CLEAN_HPTE(_hpte) ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY)) | |
1208 | #define DIRTY_HPTE(_hpte) ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY)) | |
1209 | ||
715c5407 DG |
1210 | /* |
1211 | * Get the fd to access the kernel htab, re-opening it if necessary | |
1212 | */ | |
1213 | static int get_htab_fd(sPAPRMachineState *spapr) | |
1214 | { | |
1215 | if (spapr->htab_fd >= 0) { | |
1216 | return spapr->htab_fd; | |
1217 | } | |
1218 | ||
1219 | spapr->htab_fd = kvmppc_get_htab_fd(false); | |
1220 | if (spapr->htab_fd < 0) { | |
1221 | error_report("Unable to open fd for reading hash table from KVM: %s", | |
1222 | strerror(errno)); | |
1223 | } | |
1224 | ||
1225 | return spapr->htab_fd; | |
1226 | } | |
1227 | ||
b4db5413 | 1228 | void close_htab_fd(sPAPRMachineState *spapr) |
715c5407 DG |
1229 | { |
1230 | if (spapr->htab_fd >= 0) { | |
1231 | close(spapr->htab_fd); | |
1232 | } | |
1233 | spapr->htab_fd = -1; | |
1234 | } | |
1235 | ||
e57ca75c DG |
1236 | static hwaddr spapr_hpt_mask(PPCVirtualHypervisor *vhyp) |
1237 | { | |
1238 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1239 | ||
1240 | return HTAB_SIZE(spapr) / HASH_PTEG_SIZE_64 - 1; | |
1241 | } | |
1242 | ||
1243 | static const ppc_hash_pte64_t *spapr_map_hptes(PPCVirtualHypervisor *vhyp, | |
1244 | hwaddr ptex, int n) | |
1245 | { | |
1246 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1247 | hwaddr pte_offset = ptex * HASH_PTE_SIZE_64; | |
1248 | ||
1249 | if (!spapr->htab) { | |
1250 | /* | |
1251 | * HTAB is controlled by KVM. Fetch into temporary buffer | |
1252 | */ | |
1253 | ppc_hash_pte64_t *hptes = g_malloc(n * HASH_PTE_SIZE_64); | |
1254 | kvmppc_read_hptes(hptes, ptex, n); | |
1255 | return hptes; | |
1256 | } | |
1257 | ||
1258 | /* | |
1259 | * HTAB is controlled by QEMU. Just point to the internally | |
1260 | * accessible PTEG. | |
1261 | */ | |
1262 | return (const ppc_hash_pte64_t *)(spapr->htab + pte_offset); | |
1263 | } | |
1264 | ||
1265 | static void spapr_unmap_hptes(PPCVirtualHypervisor *vhyp, | |
1266 | const ppc_hash_pte64_t *hptes, | |
1267 | hwaddr ptex, int n) | |
1268 | { | |
1269 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1270 | ||
1271 | if (!spapr->htab) { | |
1272 | g_free((void *)hptes); | |
1273 | } | |
1274 | ||
1275 | /* Nothing to do for qemu managed HPT */ | |
1276 | } | |
1277 | ||
1278 | static void spapr_store_hpte(PPCVirtualHypervisor *vhyp, hwaddr ptex, | |
1279 | uint64_t pte0, uint64_t pte1) | |
1280 | { | |
1281 | sPAPRMachineState *spapr = SPAPR_MACHINE(vhyp); | |
1282 | hwaddr offset = ptex * HASH_PTE_SIZE_64; | |
1283 | ||
1284 | if (!spapr->htab) { | |
1285 | kvmppc_write_hpte(ptex, pte0, pte1); | |
1286 | } else { | |
1287 | stq_p(spapr->htab + offset, pte0); | |
1288 | stq_p(spapr->htab + offset + HASH_PTE_SIZE_64 / 2, pte1); | |
1289 | } | |
1290 | } | |
1291 | ||
0b0b8310 | 1292 | int spapr_hpt_shift_for_ramsize(uint64_t ramsize) |
8dfe8e7f DG |
1293 | { |
1294 | int shift; | |
1295 | ||
1296 | /* We aim for a hash table of size 1/128 the size of RAM (rounded | |
1297 | * up). The PAPR recommendation is actually 1/64 of RAM size, but | |
1298 | * that's much more than is needed for Linux guests */ | |
1299 | shift = ctz64(pow2ceil(ramsize)) - 7; | |
1300 | shift = MAX(shift, 18); /* Minimum architected size */ | |
1301 | shift = MIN(shift, 46); /* Maximum architected size */ | |
1302 | return shift; | |
1303 | } | |
1304 | ||
06ec79e8 BR |
1305 | void spapr_free_hpt(sPAPRMachineState *spapr) |
1306 | { | |
1307 | g_free(spapr->htab); | |
1308 | spapr->htab = NULL; | |
1309 | spapr->htab_shift = 0; | |
1310 | close_htab_fd(spapr); | |
1311 | } | |
1312 | ||
2772cf6b DG |
1313 | void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift, |
1314 | Error **errp) | |
7f763a5d | 1315 | { |
c5f54f3e DG |
1316 | long rc; |
1317 | ||
1318 | /* Clean up any HPT info from a previous boot */ | |
06ec79e8 | 1319 | spapr_free_hpt(spapr); |
c5f54f3e DG |
1320 | |
1321 | rc = kvmppc_reset_htab(shift); | |
1322 | if (rc < 0) { | |
1323 | /* kernel-side HPT needed, but couldn't allocate one */ | |
1324 | error_setg_errno(errp, errno, | |
1325 | "Failed to allocate KVM HPT of order %d (try smaller maxmem?)", | |
1326 | shift); | |
1327 | /* This is almost certainly fatal, but if the caller really | |
1328 | * wants to carry on with shift == 0, it's welcome to try */ | |
1329 | } else if (rc > 0) { | |
1330 | /* kernel-side HPT allocated */ | |
1331 | if (rc != shift) { | |
1332 | error_setg(errp, | |
1333 | "Requested order %d HPT, but kernel allocated order %ld (try smaller maxmem?)", | |
1334 | shift, rc); | |
7735feda BR |
1335 | } |
1336 | ||
7f763a5d | 1337 | spapr->htab_shift = shift; |
c18ad9a5 | 1338 | spapr->htab = NULL; |
b817772a | 1339 | } else { |
c5f54f3e DG |
1340 | /* kernel-side HPT not needed, allocate in userspace instead */ |
1341 | size_t size = 1ULL << shift; | |
1342 | int i; | |
b817772a | 1343 | |
c5f54f3e DG |
1344 | spapr->htab = qemu_memalign(size, size); |
1345 | if (!spapr->htab) { | |
1346 | error_setg_errno(errp, errno, | |
1347 | "Could not allocate HPT of order %d", shift); | |
1348 | return; | |
7735feda BR |
1349 | } |
1350 | ||
c5f54f3e DG |
1351 | memset(spapr->htab, 0, size); |
1352 | spapr->htab_shift = shift; | |
e6b8fd24 | 1353 | |
c5f54f3e DG |
1354 | for (i = 0; i < size / HASH_PTE_SIZE_64; i++) { |
1355 | DIRTY_HPTE(HPTE(spapr->htab, i)); | |
e6b8fd24 | 1356 | } |
7f763a5d | 1357 | } |
9fdf0c29 DG |
1358 | } |
1359 | ||
b4db5413 SJS |
1360 | void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr) |
1361 | { | |
2772cf6b DG |
1362 | int hpt_shift; |
1363 | ||
1364 | if ((spapr->resize_hpt == SPAPR_RESIZE_HPT_DISABLED) | |
1365 | || (spapr->cas_reboot | |
1366 | && !spapr_ovec_test(spapr->ov5_cas, OV5_HPT_RESIZE))) { | |
1367 | hpt_shift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->maxram_size); | |
1368 | } else { | |
1369 | hpt_shift = spapr_hpt_shift_for_ramsize(MACHINE(spapr)->ram_size); | |
1370 | } | |
1371 | spapr_reallocate_hpt(spapr, hpt_shift, &error_fatal); | |
1372 | ||
b4db5413 SJS |
1373 | if (spapr->vrma_adjust) { |
1374 | spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), | |
1375 | spapr->htab_shift); | |
1376 | } | |
1377 | /* We're setting up a hash table, so that means we're not radix */ | |
1378 | spapr->patb_entry = 0; | |
1379 | } | |
1380 | ||
4f01a637 | 1381 | static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque) |
9e3f9733 AG |
1382 | { |
1383 | bool matched = false; | |
1384 | ||
1385 | if (object_dynamic_cast(OBJECT(sbdev), TYPE_SPAPR_PCI_HOST_BRIDGE)) { | |
1386 | matched = true; | |
1387 | } | |
1388 | ||
1389 | if (!matched) { | |
1390 | error_report("Device %s is not supported by this machine yet.", | |
1391 | qdev_fw_name(DEVICE(sbdev))); | |
1392 | exit(1); | |
1393 | } | |
9e3f9733 AG |
1394 | } |
1395 | ||
c8787ad4 | 1396 | static void ppc_spapr_reset(void) |
a3467baa | 1397 | { |
c5f54f3e DG |
1398 | MachineState *machine = MACHINE(qdev_get_machine()); |
1399 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); | |
182735ef | 1400 | PowerPCCPU *first_ppc_cpu; |
b7d1f77a | 1401 | uint32_t rtas_limit; |
cae172ab | 1402 | hwaddr rtas_addr, fdt_addr; |
997b6cfc DG |
1403 | void *fdt; |
1404 | int rc; | |
259186a7 | 1405 | |
9e3f9733 AG |
1406 | /* Check for unknown sysbus devices */ |
1407 | foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL); | |
1408 | ||
b4db5413 SJS |
1409 | if (kvm_enabled() && kvmppc_has_cap_mmu_radix()) { |
1410 | /* If using KVM with radix mode available, VCPUs can be started | |
1411 | * without a HPT because KVM will start them in radix mode. | |
1412 | * Set the GR bit in PATB so that we know there is no HPT. */ | |
1413 | spapr->patb_entry = PATBE1_GR; | |
1414 | } else { | |
b4db5413 | 1415 | spapr_setup_hpt_and_vrma(spapr); |
c5f54f3e | 1416 | } |
a3467baa | 1417 | |
c8787ad4 | 1418 | qemu_devices_reset(); |
56258174 | 1419 | spapr_clear_pending_events(spapr); |
a3467baa | 1420 | |
b7d1f77a BH |
1421 | /* |
1422 | * We place the device tree and RTAS just below either the top of the RMA, | |
1423 | * or just below 2GB, whichever is lowere, so that it can be | |
1424 | * processed with 32-bit real mode code if necessary | |
1425 | */ | |
1426 | rtas_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR); | |
cae172ab DG |
1427 | rtas_addr = rtas_limit - RTAS_MAX_SIZE; |
1428 | fdt_addr = rtas_addr - FDT_MAX_SIZE; | |
b7d1f77a | 1429 | |
6787d27b MR |
1430 | /* if this reset wasn't generated by CAS, we should reset our |
1431 | * negotiated options and start from scratch */ | |
1432 | if (!spapr->cas_reboot) { | |
1433 | spapr_ovec_cleanup(spapr->ov5_cas); | |
1434 | spapr->ov5_cas = spapr_ovec_new(); | |
66d5c492 DG |
1435 | |
1436 | ppc_set_compat_all(spapr->max_compat_pvr, &error_fatal); | |
6787d27b MR |
1437 | } |
1438 | ||
cae172ab | 1439 | fdt = spapr_build_fdt(spapr, rtas_addr, spapr->rtas_size); |
a3467baa | 1440 | |
2cac78c1 | 1441 | spapr_load_rtas(spapr, fdt, rtas_addr); |
b7d1f77a | 1442 | |
997b6cfc DG |
1443 | rc = fdt_pack(fdt); |
1444 | ||
1445 | /* Should only fail if we've built a corrupted tree */ | |
1446 | assert(rc == 0); | |
1447 | ||
1448 | if (fdt_totalsize(fdt) > FDT_MAX_SIZE) { | |
1449 | error_report("FDT too big ! 0x%x bytes (max is 0x%x)", | |
1450 | fdt_totalsize(fdt), FDT_MAX_SIZE); | |
1451 | exit(1); | |
1452 | } | |
1453 | ||
1454 | /* Load the fdt */ | |
1455 | qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt)); | |
cae172ab | 1456 | cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt)); |
997b6cfc DG |
1457 | g_free(fdt); |
1458 | ||
a3467baa | 1459 | /* Set up the entry state */ |
182735ef | 1460 | first_ppc_cpu = POWERPC_CPU(first_cpu); |
cae172ab | 1461 | first_ppc_cpu->env.gpr[3] = fdt_addr; |
182735ef AF |
1462 | first_ppc_cpu->env.gpr[5] = 0; |
1463 | first_cpu->halted = 0; | |
1b718907 | 1464 | first_ppc_cpu->env.nip = SPAPR_ENTRY_POINT; |
a3467baa | 1465 | |
6787d27b | 1466 | spapr->cas_reboot = false; |
a3467baa DG |
1467 | } |
1468 | ||
28e02042 | 1469 | static void spapr_create_nvram(sPAPRMachineState *spapr) |
639e8102 | 1470 | { |
2ff3de68 | 1471 | DeviceState *dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram"); |
3978b863 | 1472 | DriveInfo *dinfo = drive_get(IF_PFLASH, 0, 0); |
639e8102 | 1473 | |
3978b863 | 1474 | if (dinfo) { |
6231a6da MA |
1475 | qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo), |
1476 | &error_fatal); | |
639e8102 DG |
1477 | } |
1478 | ||
1479 | qdev_init_nofail(dev); | |
1480 | ||
1481 | spapr->nvram = (struct sPAPRNVRAM *)dev; | |
1482 | } | |
1483 | ||
28e02042 | 1484 | static void spapr_rtc_create(sPAPRMachineState *spapr) |
28df36a1 | 1485 | { |
147ff807 CLG |
1486 | object_initialize(&spapr->rtc, sizeof(spapr->rtc), TYPE_SPAPR_RTC); |
1487 | object_property_add_child(OBJECT(spapr), "rtc", OBJECT(&spapr->rtc), | |
1488 | &error_fatal); | |
1489 | object_property_set_bool(OBJECT(&spapr->rtc), true, "realized", | |
1490 | &error_fatal); | |
1491 | object_property_add_alias(OBJECT(spapr), "rtc-time", OBJECT(&spapr->rtc), | |
1492 | "date", &error_fatal); | |
28df36a1 DG |
1493 | } |
1494 | ||
8c57b867 | 1495 | /* Returns whether we want to use VGA or not */ |
14c6a894 | 1496 | static bool spapr_vga_init(PCIBus *pci_bus, Error **errp) |
f28359d8 | 1497 | { |
8c57b867 | 1498 | switch (vga_interface_type) { |
8c57b867 | 1499 | case VGA_NONE: |
7effdaa3 MW |
1500 | return false; |
1501 | case VGA_DEVICE: | |
1502 | return true; | |
1ddcae82 | 1503 | case VGA_STD: |
b798c190 | 1504 | case VGA_VIRTIO: |
1ddcae82 | 1505 | return pci_vga_init(pci_bus) != NULL; |
8c57b867 | 1506 | default: |
14c6a894 DG |
1507 | error_setg(errp, |
1508 | "Unsupported VGA mode, only -vga std or -vga virtio is supported"); | |
1509 | return false; | |
f28359d8 | 1510 | } |
f28359d8 LZ |
1511 | } |
1512 | ||
880ae7de DG |
1513 | static int spapr_post_load(void *opaque, int version_id) |
1514 | { | |
28e02042 | 1515 | sPAPRMachineState *spapr = (sPAPRMachineState *)opaque; |
880ae7de DG |
1516 | int err = 0; |
1517 | ||
a7ff1212 | 1518 | if (!object_dynamic_cast(OBJECT(spapr->ics), TYPE_ICS_KVM)) { |
5bc8d26d CLG |
1519 | CPUState *cs; |
1520 | CPU_FOREACH(cs) { | |
1521 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
1522 | icp_resend(ICP(cpu->intc)); | |
a7ff1212 CLG |
1523 | } |
1524 | } | |
1525 | ||
631b22ea | 1526 | /* In earlier versions, there was no separate qdev for the PAPR |
880ae7de DG |
1527 | * RTC, so the RTC offset was stored directly in sPAPREnvironment. |
1528 | * So when migrating from those versions, poke the incoming offset | |
1529 | * value into the RTC device */ | |
1530 | if (version_id < 3) { | |
147ff807 | 1531 | err = spapr_rtc_import_offset(&spapr->rtc, spapr->rtc_offset); |
880ae7de DG |
1532 | } |
1533 | ||
d39c90f5 BR |
1534 | if (spapr->patb_entry) { |
1535 | PowerPCCPU *cpu = POWERPC_CPU(first_cpu); | |
1536 | bool radix = !!(spapr->patb_entry & PATBE1_GR); | |
1537 | bool gtse = !!(cpu->env.spr[SPR_LPCR] & LPCR_GTSE); | |
1538 | ||
1539 | err = kvmppc_configure_v3_mmu(cpu, radix, gtse, spapr->patb_entry); | |
1540 | if (err) { | |
1541 | error_report("Process table config unsupported by the host"); | |
1542 | return -EINVAL; | |
1543 | } | |
1544 | } | |
1545 | ||
880ae7de DG |
1546 | return err; |
1547 | } | |
1548 | ||
1549 | static bool version_before_3(void *opaque, int version_id) | |
1550 | { | |
1551 | return version_id < 3; | |
1552 | } | |
1553 | ||
fd38804b DHB |
1554 | static bool spapr_pending_events_needed(void *opaque) |
1555 | { | |
1556 | sPAPRMachineState *spapr = (sPAPRMachineState *)opaque; | |
1557 | return !QTAILQ_EMPTY(&spapr->pending_events); | |
1558 | } | |
1559 | ||
1560 | static const VMStateDescription vmstate_spapr_event_entry = { | |
1561 | .name = "spapr_event_log_entry", | |
1562 | .version_id = 1, | |
1563 | .minimum_version_id = 1, | |
1564 | .fields = (VMStateField[]) { | |
5341258e DG |
1565 | VMSTATE_UINT32(summary, sPAPREventLogEntry), |
1566 | VMSTATE_UINT32(extended_length, sPAPREventLogEntry), | |
fd38804b | 1567 | VMSTATE_VBUFFER_ALLOC_UINT32(extended_log, sPAPREventLogEntry, 0, |
5341258e | 1568 | NULL, extended_length), |
fd38804b DHB |
1569 | VMSTATE_END_OF_LIST() |
1570 | }, | |
1571 | }; | |
1572 | ||
1573 | static const VMStateDescription vmstate_spapr_pending_events = { | |
1574 | .name = "spapr_pending_events", | |
1575 | .version_id = 1, | |
1576 | .minimum_version_id = 1, | |
1577 | .needed = spapr_pending_events_needed, | |
1578 | .fields = (VMStateField[]) { | |
1579 | VMSTATE_QTAILQ_V(pending_events, sPAPRMachineState, 1, | |
1580 | vmstate_spapr_event_entry, sPAPREventLogEntry, next), | |
1581 | VMSTATE_END_OF_LIST() | |
1582 | }, | |
1583 | }; | |
1584 | ||
62ef3760 MR |
1585 | static bool spapr_ov5_cas_needed(void *opaque) |
1586 | { | |
1587 | sPAPRMachineState *spapr = opaque; | |
1588 | sPAPROptionVector *ov5_mask = spapr_ovec_new(); | |
1589 | sPAPROptionVector *ov5_legacy = spapr_ovec_new(); | |
1590 | sPAPROptionVector *ov5_removed = spapr_ovec_new(); | |
1591 | bool cas_needed; | |
1592 | ||
1593 | /* Prior to the introduction of sPAPROptionVector, we had two option | |
1594 | * vectors we dealt with: OV5_FORM1_AFFINITY, and OV5_DRCONF_MEMORY. | |
1595 | * Both of these options encode machine topology into the device-tree | |
1596 | * in such a way that the now-booted OS should still be able to interact | |
1597 | * appropriately with QEMU regardless of what options were actually | |
1598 | * negotiatied on the source side. | |
1599 | * | |
1600 | * As such, we can avoid migrating the CAS-negotiated options if these | |
1601 | * are the only options available on the current machine/platform. | |
1602 | * Since these are the only options available for pseries-2.7 and | |
1603 | * earlier, this allows us to maintain old->new/new->old migration | |
1604 | * compatibility. | |
1605 | * | |
1606 | * For QEMU 2.8+, there are additional CAS-negotiatable options available | |
1607 | * via default pseries-2.8 machines and explicit command-line parameters. | |
1608 | * Some of these options, like OV5_HP_EVT, *do* require QEMU to be aware | |
1609 | * of the actual CAS-negotiated values to continue working properly. For | |
1610 | * example, availability of memory unplug depends on knowing whether | |
1611 | * OV5_HP_EVT was negotiated via CAS. | |
1612 | * | |
1613 | * Thus, for any cases where the set of available CAS-negotiatable | |
1614 | * options extends beyond OV5_FORM1_AFFINITY and OV5_DRCONF_MEMORY, we | |
1615 | * include the CAS-negotiated options in the migration stream. | |
1616 | */ | |
1617 | spapr_ovec_set(ov5_mask, OV5_FORM1_AFFINITY); | |
1618 | spapr_ovec_set(ov5_mask, OV5_DRCONF_MEMORY); | |
1619 | ||
1620 | /* spapr_ovec_diff returns true if bits were removed. we avoid using | |
1621 | * the mask itself since in the future it's possible "legacy" bits may be | |
1622 | * removed via machine options, which could generate a false positive | |
1623 | * that breaks migration. | |
1624 | */ | |
1625 | spapr_ovec_intersect(ov5_legacy, spapr->ov5, ov5_mask); | |
1626 | cas_needed = spapr_ovec_diff(ov5_removed, spapr->ov5, ov5_legacy); | |
1627 | ||
1628 | spapr_ovec_cleanup(ov5_mask); | |
1629 | spapr_ovec_cleanup(ov5_legacy); | |
1630 | spapr_ovec_cleanup(ov5_removed); | |
1631 | ||
1632 | return cas_needed; | |
1633 | } | |
1634 | ||
1635 | static const VMStateDescription vmstate_spapr_ov5_cas = { | |
1636 | .name = "spapr_option_vector_ov5_cas", | |
1637 | .version_id = 1, | |
1638 | .minimum_version_id = 1, | |
1639 | .needed = spapr_ov5_cas_needed, | |
1640 | .fields = (VMStateField[]) { | |
1641 | VMSTATE_STRUCT_POINTER_V(ov5_cas, sPAPRMachineState, 1, | |
1642 | vmstate_spapr_ovec, sPAPROptionVector), | |
1643 | VMSTATE_END_OF_LIST() | |
1644 | }, | |
1645 | }; | |
1646 | ||
9861bb3e SJS |
1647 | static bool spapr_patb_entry_needed(void *opaque) |
1648 | { | |
1649 | sPAPRMachineState *spapr = opaque; | |
1650 | ||
1651 | return !!spapr->patb_entry; | |
1652 | } | |
1653 | ||
1654 | static const VMStateDescription vmstate_spapr_patb_entry = { | |
1655 | .name = "spapr_patb_entry", | |
1656 | .version_id = 1, | |
1657 | .minimum_version_id = 1, | |
1658 | .needed = spapr_patb_entry_needed, | |
1659 | .fields = (VMStateField[]) { | |
1660 | VMSTATE_UINT64(patb_entry, sPAPRMachineState), | |
1661 | VMSTATE_END_OF_LIST() | |
1662 | }, | |
1663 | }; | |
1664 | ||
4be21d56 DG |
1665 | static const VMStateDescription vmstate_spapr = { |
1666 | .name = "spapr", | |
880ae7de | 1667 | .version_id = 3, |
4be21d56 | 1668 | .minimum_version_id = 1, |
880ae7de | 1669 | .post_load = spapr_post_load, |
3aff6c2f | 1670 | .fields = (VMStateField[]) { |
880ae7de DG |
1671 | /* used to be @next_irq */ |
1672 | VMSTATE_UNUSED_BUFFER(version_before_3, 0, 4), | |
4be21d56 DG |
1673 | |
1674 | /* RTC offset */ | |
28e02042 | 1675 | VMSTATE_UINT64_TEST(rtc_offset, sPAPRMachineState, version_before_3), |
880ae7de | 1676 | |
28e02042 | 1677 | VMSTATE_PPC_TIMEBASE_V(tb, sPAPRMachineState, 2), |
4be21d56 DG |
1678 | VMSTATE_END_OF_LIST() |
1679 | }, | |
62ef3760 MR |
1680 | .subsections = (const VMStateDescription*[]) { |
1681 | &vmstate_spapr_ov5_cas, | |
9861bb3e | 1682 | &vmstate_spapr_patb_entry, |
fd38804b | 1683 | &vmstate_spapr_pending_events, |
62ef3760 MR |
1684 | NULL |
1685 | } | |
4be21d56 DG |
1686 | }; |
1687 | ||
4be21d56 DG |
1688 | static int htab_save_setup(QEMUFile *f, void *opaque) |
1689 | { | |
28e02042 | 1690 | sPAPRMachineState *spapr = opaque; |
4be21d56 | 1691 | |
4be21d56 | 1692 | /* "Iteration" header */ |
3a384297 BR |
1693 | if (!spapr->htab_shift) { |
1694 | qemu_put_be32(f, -1); | |
1695 | } else { | |
1696 | qemu_put_be32(f, spapr->htab_shift); | |
1697 | } | |
4be21d56 | 1698 | |
e68cb8b4 AK |
1699 | if (spapr->htab) { |
1700 | spapr->htab_save_index = 0; | |
1701 | spapr->htab_first_pass = true; | |
1702 | } else { | |
3a384297 BR |
1703 | if (spapr->htab_shift) { |
1704 | assert(kvm_enabled()); | |
1705 | } | |
e68cb8b4 AK |
1706 | } |
1707 | ||
1708 | ||
4be21d56 DG |
1709 | return 0; |
1710 | } | |
1711 | ||
28e02042 | 1712 | static void htab_save_first_pass(QEMUFile *f, sPAPRMachineState *spapr, |
4be21d56 DG |
1713 | int64_t max_ns) |
1714 | { | |
378bc217 | 1715 | bool has_timeout = max_ns != -1; |
4be21d56 DG |
1716 | int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; |
1717 | int index = spapr->htab_save_index; | |
bc72ad67 | 1718 | int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
4be21d56 DG |
1719 | |
1720 | assert(spapr->htab_first_pass); | |
1721 | ||
1722 | do { | |
1723 | int chunkstart; | |
1724 | ||
1725 | /* Consume invalid HPTEs */ | |
1726 | while ((index < htabslots) | |
1727 | && !HPTE_VALID(HPTE(spapr->htab, index))) { | |
4be21d56 | 1728 | CLEAN_HPTE(HPTE(spapr->htab, index)); |
24ec2863 | 1729 | index++; |
4be21d56 DG |
1730 | } |
1731 | ||
1732 | /* Consume valid HPTEs */ | |
1733 | chunkstart = index; | |
338c25b6 | 1734 | while ((index < htabslots) && (index - chunkstart < USHRT_MAX) |
4be21d56 | 1735 | && HPTE_VALID(HPTE(spapr->htab, index))) { |
4be21d56 | 1736 | CLEAN_HPTE(HPTE(spapr->htab, index)); |
24ec2863 | 1737 | index++; |
4be21d56 DG |
1738 | } |
1739 | ||
1740 | if (index > chunkstart) { | |
1741 | int n_valid = index - chunkstart; | |
1742 | ||
1743 | qemu_put_be32(f, chunkstart); | |
1744 | qemu_put_be16(f, n_valid); | |
1745 | qemu_put_be16(f, 0); | |
1746 | qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), | |
1747 | HASH_PTE_SIZE_64 * n_valid); | |
1748 | ||
378bc217 DG |
1749 | if (has_timeout && |
1750 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { | |
4be21d56 DG |
1751 | break; |
1752 | } | |
1753 | } | |
1754 | } while ((index < htabslots) && !qemu_file_rate_limit(f)); | |
1755 | ||
1756 | if (index >= htabslots) { | |
1757 | assert(index == htabslots); | |
1758 | index = 0; | |
1759 | spapr->htab_first_pass = false; | |
1760 | } | |
1761 | spapr->htab_save_index = index; | |
1762 | } | |
1763 | ||
28e02042 | 1764 | static int htab_save_later_pass(QEMUFile *f, sPAPRMachineState *spapr, |
e68cb8b4 | 1765 | int64_t max_ns) |
4be21d56 DG |
1766 | { |
1767 | bool final = max_ns < 0; | |
1768 | int htabslots = HTAB_SIZE(spapr) / HASH_PTE_SIZE_64; | |
1769 | int examined = 0, sent = 0; | |
1770 | int index = spapr->htab_save_index; | |
bc72ad67 | 1771 | int64_t starttime = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
4be21d56 DG |
1772 | |
1773 | assert(!spapr->htab_first_pass); | |
1774 | ||
1775 | do { | |
1776 | int chunkstart, invalidstart; | |
1777 | ||
1778 | /* Consume non-dirty HPTEs */ | |
1779 | while ((index < htabslots) | |
1780 | && !HPTE_DIRTY(HPTE(spapr->htab, index))) { | |
1781 | index++; | |
1782 | examined++; | |
1783 | } | |
1784 | ||
1785 | chunkstart = index; | |
1786 | /* Consume valid dirty HPTEs */ | |
338c25b6 | 1787 | while ((index < htabslots) && (index - chunkstart < USHRT_MAX) |
4be21d56 DG |
1788 | && HPTE_DIRTY(HPTE(spapr->htab, index)) |
1789 | && HPTE_VALID(HPTE(spapr->htab, index))) { | |
1790 | CLEAN_HPTE(HPTE(spapr->htab, index)); | |
1791 | index++; | |
1792 | examined++; | |
1793 | } | |
1794 | ||
1795 | invalidstart = index; | |
1796 | /* Consume invalid dirty HPTEs */ | |
338c25b6 | 1797 | while ((index < htabslots) && (index - invalidstart < USHRT_MAX) |
4be21d56 DG |
1798 | && HPTE_DIRTY(HPTE(spapr->htab, index)) |
1799 | && !HPTE_VALID(HPTE(spapr->htab, index))) { | |
1800 | CLEAN_HPTE(HPTE(spapr->htab, index)); | |
1801 | index++; | |
1802 | examined++; | |
1803 | } | |
1804 | ||
1805 | if (index > chunkstart) { | |
1806 | int n_valid = invalidstart - chunkstart; | |
1807 | int n_invalid = index - invalidstart; | |
1808 | ||
1809 | qemu_put_be32(f, chunkstart); | |
1810 | qemu_put_be16(f, n_valid); | |
1811 | qemu_put_be16(f, n_invalid); | |
1812 | qemu_put_buffer(f, HPTE(spapr->htab, chunkstart), | |
1813 | HASH_PTE_SIZE_64 * n_valid); | |
1814 | sent += index - chunkstart; | |
1815 | ||
bc72ad67 | 1816 | if (!final && (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - starttime) > max_ns) { |
4be21d56 DG |
1817 | break; |
1818 | } | |
1819 | } | |
1820 | ||
1821 | if (examined >= htabslots) { | |
1822 | break; | |
1823 | } | |
1824 | ||
1825 | if (index >= htabslots) { | |
1826 | assert(index == htabslots); | |
1827 | index = 0; | |
1828 | } | |
1829 | } while ((examined < htabslots) && (!qemu_file_rate_limit(f) || final)); | |
1830 | ||
1831 | if (index >= htabslots) { | |
1832 | assert(index == htabslots); | |
1833 | index = 0; | |
1834 | } | |
1835 | ||
1836 | spapr->htab_save_index = index; | |
1837 | ||
e68cb8b4 | 1838 | return (examined >= htabslots) && (sent == 0) ? 1 : 0; |
4be21d56 DG |
1839 | } |
1840 | ||
e68cb8b4 AK |
1841 | #define MAX_ITERATION_NS 5000000 /* 5 ms */ |
1842 | #define MAX_KVM_BUF_SIZE 2048 | |
1843 | ||
4be21d56 DG |
1844 | static int htab_save_iterate(QEMUFile *f, void *opaque) |
1845 | { | |
28e02042 | 1846 | sPAPRMachineState *spapr = opaque; |
715c5407 | 1847 | int fd; |
e68cb8b4 | 1848 | int rc = 0; |
4be21d56 DG |
1849 | |
1850 | /* Iteration header */ | |
3a384297 BR |
1851 | if (!spapr->htab_shift) { |
1852 | qemu_put_be32(f, -1); | |
e8cd4247 | 1853 | return 1; |
3a384297 BR |
1854 | } else { |
1855 | qemu_put_be32(f, 0); | |
1856 | } | |
4be21d56 | 1857 | |
e68cb8b4 AK |
1858 | if (!spapr->htab) { |
1859 | assert(kvm_enabled()); | |
1860 | ||
715c5407 DG |
1861 | fd = get_htab_fd(spapr); |
1862 | if (fd < 0) { | |
1863 | return fd; | |
01a57972 SMJ |
1864 | } |
1865 | ||
715c5407 | 1866 | rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, MAX_ITERATION_NS); |
e68cb8b4 AK |
1867 | if (rc < 0) { |
1868 | return rc; | |
1869 | } | |
1870 | } else if (spapr->htab_first_pass) { | |
4be21d56 DG |
1871 | htab_save_first_pass(f, spapr, MAX_ITERATION_NS); |
1872 | } else { | |
e68cb8b4 | 1873 | rc = htab_save_later_pass(f, spapr, MAX_ITERATION_NS); |
4be21d56 DG |
1874 | } |
1875 | ||
1876 | /* End marker */ | |
1877 | qemu_put_be32(f, 0); | |
1878 | qemu_put_be16(f, 0); | |
1879 | qemu_put_be16(f, 0); | |
1880 | ||
e68cb8b4 | 1881 | return rc; |
4be21d56 DG |
1882 | } |
1883 | ||
1884 | static int htab_save_complete(QEMUFile *f, void *opaque) | |
1885 | { | |
28e02042 | 1886 | sPAPRMachineState *spapr = opaque; |
715c5407 | 1887 | int fd; |
4be21d56 DG |
1888 | |
1889 | /* Iteration header */ | |
3a384297 BR |
1890 | if (!spapr->htab_shift) { |
1891 | qemu_put_be32(f, -1); | |
1892 | return 0; | |
1893 | } else { | |
1894 | qemu_put_be32(f, 0); | |
1895 | } | |
4be21d56 | 1896 | |
e68cb8b4 AK |
1897 | if (!spapr->htab) { |
1898 | int rc; | |
1899 | ||
1900 | assert(kvm_enabled()); | |
1901 | ||
715c5407 DG |
1902 | fd = get_htab_fd(spapr); |
1903 | if (fd < 0) { | |
1904 | return fd; | |
01a57972 SMJ |
1905 | } |
1906 | ||
715c5407 | 1907 | rc = kvmppc_save_htab(f, fd, MAX_KVM_BUF_SIZE, -1); |
e68cb8b4 AK |
1908 | if (rc < 0) { |
1909 | return rc; | |
1910 | } | |
e68cb8b4 | 1911 | } else { |
378bc217 DG |
1912 | if (spapr->htab_first_pass) { |
1913 | htab_save_first_pass(f, spapr, -1); | |
1914 | } | |
e68cb8b4 AK |
1915 | htab_save_later_pass(f, spapr, -1); |
1916 | } | |
4be21d56 DG |
1917 | |
1918 | /* End marker */ | |
1919 | qemu_put_be32(f, 0); | |
1920 | qemu_put_be16(f, 0); | |
1921 | qemu_put_be16(f, 0); | |
1922 | ||
1923 | return 0; | |
1924 | } | |
1925 | ||
1926 | static int htab_load(QEMUFile *f, void *opaque, int version_id) | |
1927 | { | |
28e02042 | 1928 | sPAPRMachineState *spapr = opaque; |
4be21d56 | 1929 | uint32_t section_hdr; |
e68cb8b4 | 1930 | int fd = -1; |
4be21d56 DG |
1931 | |
1932 | if (version_id < 1 || version_id > 1) { | |
98a5d100 | 1933 | error_report("htab_load() bad version"); |
4be21d56 DG |
1934 | return -EINVAL; |
1935 | } | |
1936 | ||
1937 | section_hdr = qemu_get_be32(f); | |
1938 | ||
3a384297 BR |
1939 | if (section_hdr == -1) { |
1940 | spapr_free_hpt(spapr); | |
1941 | return 0; | |
1942 | } | |
1943 | ||
4be21d56 | 1944 | if (section_hdr) { |
9897e462 | 1945 | Error *local_err = NULL; |
c5f54f3e DG |
1946 | |
1947 | /* First section gives the htab size */ | |
1948 | spapr_reallocate_hpt(spapr, section_hdr, &local_err); | |
1949 | if (local_err) { | |
1950 | error_report_err(local_err); | |
4be21d56 DG |
1951 | return -EINVAL; |
1952 | } | |
1953 | return 0; | |
1954 | } | |
1955 | ||
e68cb8b4 AK |
1956 | if (!spapr->htab) { |
1957 | assert(kvm_enabled()); | |
1958 | ||
1959 | fd = kvmppc_get_htab_fd(true); | |
1960 | if (fd < 0) { | |
98a5d100 DG |
1961 | error_report("Unable to open fd to restore KVM hash table: %s", |
1962 | strerror(errno)); | |
e68cb8b4 AK |
1963 | } |
1964 | } | |
1965 | ||
4be21d56 DG |
1966 | while (true) { |
1967 | uint32_t index; | |
1968 | uint16_t n_valid, n_invalid; | |
1969 | ||
1970 | index = qemu_get_be32(f); | |
1971 | n_valid = qemu_get_be16(f); | |
1972 | n_invalid = qemu_get_be16(f); | |
1973 | ||
1974 | if ((index == 0) && (n_valid == 0) && (n_invalid == 0)) { | |
1975 | /* End of Stream */ | |
1976 | break; | |
1977 | } | |
1978 | ||
e68cb8b4 | 1979 | if ((index + n_valid + n_invalid) > |
4be21d56 DG |
1980 | (HTAB_SIZE(spapr) / HASH_PTE_SIZE_64)) { |
1981 | /* Bad index in stream */ | |
98a5d100 DG |
1982 | error_report( |
1983 | "htab_load() bad index %d (%hd+%hd entries) in htab stream (htab_shift=%d)", | |
1984 | index, n_valid, n_invalid, spapr->htab_shift); | |
4be21d56 DG |
1985 | return -EINVAL; |
1986 | } | |
1987 | ||
e68cb8b4 AK |
1988 | if (spapr->htab) { |
1989 | if (n_valid) { | |
1990 | qemu_get_buffer(f, HPTE(spapr->htab, index), | |
1991 | HASH_PTE_SIZE_64 * n_valid); | |
1992 | } | |
1993 | if (n_invalid) { | |
1994 | memset(HPTE(spapr->htab, index + n_valid), 0, | |
1995 | HASH_PTE_SIZE_64 * n_invalid); | |
1996 | } | |
1997 | } else { | |
1998 | int rc; | |
1999 | ||
2000 | assert(fd >= 0); | |
2001 | ||
2002 | rc = kvmppc_load_htab_chunk(f, fd, index, n_valid, n_invalid); | |
2003 | if (rc < 0) { | |
2004 | return rc; | |
2005 | } | |
4be21d56 DG |
2006 | } |
2007 | } | |
2008 | ||
e68cb8b4 AK |
2009 | if (!spapr->htab) { |
2010 | assert(fd >= 0); | |
2011 | close(fd); | |
2012 | } | |
2013 | ||
4be21d56 DG |
2014 | return 0; |
2015 | } | |
2016 | ||
70f794fc | 2017 | static void htab_save_cleanup(void *opaque) |
c573fc03 TH |
2018 | { |
2019 | sPAPRMachineState *spapr = opaque; | |
2020 | ||
2021 | close_htab_fd(spapr); | |
2022 | } | |
2023 | ||
4be21d56 | 2024 | static SaveVMHandlers savevm_htab_handlers = { |
9907e842 | 2025 | .save_setup = htab_save_setup, |
4be21d56 | 2026 | .save_live_iterate = htab_save_iterate, |
a3e06c3d | 2027 | .save_live_complete_precopy = htab_save_complete, |
70f794fc | 2028 | .save_cleanup = htab_save_cleanup, |
4be21d56 DG |
2029 | .load_state = htab_load, |
2030 | }; | |
2031 | ||
5b2128d2 AG |
2032 | static void spapr_boot_set(void *opaque, const char *boot_device, |
2033 | Error **errp) | |
2034 | { | |
2035 | MachineState *machine = MACHINE(qdev_get_machine()); | |
2036 | machine->boot_order = g_strdup(boot_device); | |
2037 | } | |
2038 | ||
224245bf DG |
2039 | static void spapr_create_lmb_dr_connectors(sPAPRMachineState *spapr) |
2040 | { | |
2041 | MachineState *machine = MACHINE(spapr); | |
2042 | uint64_t lmb_size = SPAPR_MEMORY_BLOCK_SIZE; | |
e8f986fc | 2043 | uint32_t nr_lmbs = (machine->maxram_size - machine->ram_size)/lmb_size; |
224245bf DG |
2044 | int i; |
2045 | ||
2046 | for (i = 0; i < nr_lmbs; i++) { | |
224245bf DG |
2047 | uint64_t addr; |
2048 | ||
e8f986fc | 2049 | addr = i * lmb_size + spapr->hotplug_memory.base; |
6caf3ac6 DG |
2050 | spapr_dr_connector_new(OBJECT(spapr), TYPE_SPAPR_DRC_LMB, |
2051 | addr / lmb_size); | |
224245bf DG |
2052 | } |
2053 | } | |
2054 | ||
2055 | /* | |
2056 | * If RAM size, maxmem size and individual node mem sizes aren't aligned | |
2057 | * to SPAPR_MEMORY_BLOCK_SIZE(256MB), then refuse to start the guest | |
2058 | * since we can't support such unaligned sizes with DRCONF_MEMORY. | |
2059 | */ | |
7c150d6f | 2060 | static void spapr_validate_node_memory(MachineState *machine, Error **errp) |
224245bf DG |
2061 | { |
2062 | int i; | |
2063 | ||
7c150d6f DG |
2064 | if (machine->ram_size % SPAPR_MEMORY_BLOCK_SIZE) { |
2065 | error_setg(errp, "Memory size 0x" RAM_ADDR_FMT | |
2066 | " is not aligned to %llu MiB", | |
2067 | machine->ram_size, | |
2068 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
2069 | return; | |
2070 | } | |
2071 | ||
2072 | if (machine->maxram_size % SPAPR_MEMORY_BLOCK_SIZE) { | |
2073 | error_setg(errp, "Maximum memory size 0x" RAM_ADDR_FMT | |
2074 | " is not aligned to %llu MiB", | |
2075 | machine->ram_size, | |
2076 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
2077 | return; | |
224245bf DG |
2078 | } |
2079 | ||
2080 | for (i = 0; i < nb_numa_nodes; i++) { | |
2081 | if (numa_info[i].node_mem % SPAPR_MEMORY_BLOCK_SIZE) { | |
7c150d6f DG |
2082 | error_setg(errp, |
2083 | "Node %d memory size 0x%" PRIx64 | |
2084 | " is not aligned to %llu MiB", | |
2085 | i, numa_info[i].node_mem, | |
2086 | SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
2087 | return; | |
224245bf DG |
2088 | } |
2089 | } | |
2090 | } | |
2091 | ||
535455fd IM |
2092 | /* find cpu slot in machine->possible_cpus by core_id */ |
2093 | static CPUArchId *spapr_find_cpu_slot(MachineState *ms, uint32_t id, int *idx) | |
2094 | { | |
2095 | int index = id / smp_threads; | |
2096 | ||
2097 | if (index >= ms->possible_cpus->len) { | |
2098 | return NULL; | |
2099 | } | |
2100 | if (idx) { | |
2101 | *idx = index; | |
2102 | } | |
2103 | return &ms->possible_cpus->cpus[index]; | |
2104 | } | |
2105 | ||
0c86d0fd DG |
2106 | static void spapr_init_cpus(sPAPRMachineState *spapr) |
2107 | { | |
2108 | MachineState *machine = MACHINE(spapr); | |
2109 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
2110 | char *type = spapr_get_cpu_core_type(machine->cpu_model); | |
2111 | int smt = kvmppc_smt_threads(); | |
535455fd IM |
2112 | const CPUArchIdList *possible_cpus; |
2113 | int boot_cores_nr = smp_cpus / smp_threads; | |
0c86d0fd DG |
2114 | int i; |
2115 | ||
2116 | if (!type) { | |
2117 | error_report("Unable to find sPAPR CPU Core definition"); | |
2118 | exit(1); | |
2119 | } | |
2120 | ||
535455fd | 2121 | possible_cpus = mc->possible_cpu_arch_ids(machine); |
c5514d0e | 2122 | if (mc->has_hotpluggable_cpus) { |
0c86d0fd DG |
2123 | if (smp_cpus % smp_threads) { |
2124 | error_report("smp_cpus (%u) must be multiple of threads (%u)", | |
2125 | smp_cpus, smp_threads); | |
2126 | exit(1); | |
2127 | } | |
2128 | if (max_cpus % smp_threads) { | |
2129 | error_report("max_cpus (%u) must be multiple of threads (%u)", | |
2130 | max_cpus, smp_threads); | |
2131 | exit(1); | |
2132 | } | |
0c86d0fd DG |
2133 | } else { |
2134 | if (max_cpus != smp_cpus) { | |
2135 | error_report("This machine version does not support CPU hotplug"); | |
2136 | exit(1); | |
2137 | } | |
535455fd | 2138 | boot_cores_nr = possible_cpus->len; |
0c86d0fd DG |
2139 | } |
2140 | ||
535455fd | 2141 | for (i = 0; i < possible_cpus->len; i++) { |
0c86d0fd DG |
2142 | int core_id = i * smp_threads; |
2143 | ||
c5514d0e | 2144 | if (mc->has_hotpluggable_cpus) { |
6caf3ac6 DG |
2145 | spapr_dr_connector_new(OBJECT(spapr), TYPE_SPAPR_DRC_CPU, |
2146 | (core_id / smp_threads) * smt); | |
0c86d0fd DG |
2147 | } |
2148 | ||
535455fd | 2149 | if (i < boot_cores_nr) { |
0c86d0fd DG |
2150 | Object *core = object_new(type); |
2151 | int nr_threads = smp_threads; | |
2152 | ||
2153 | /* Handle the partially filled core for older machine types */ | |
2154 | if ((i + 1) * smp_threads >= smp_cpus) { | |
2155 | nr_threads = smp_cpus - i * smp_threads; | |
2156 | } | |
2157 | ||
2158 | object_property_set_int(core, nr_threads, "nr-threads", | |
2159 | &error_fatal); | |
2160 | object_property_set_int(core, core_id, CPU_CORE_PROP_CORE_ID, | |
2161 | &error_fatal); | |
2162 | object_property_set_bool(core, true, "realized", &error_fatal); | |
2163 | } | |
2164 | } | |
2165 | g_free(type); | |
2166 | } | |
2167 | ||
9fdf0c29 | 2168 | /* pSeries LPAR / sPAPR hardware init */ |
3ef96221 | 2169 | static void ppc_spapr_init(MachineState *machine) |
9fdf0c29 | 2170 | { |
28e02042 | 2171 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
224245bf | 2172 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(machine); |
3ef96221 | 2173 | const char *kernel_filename = machine->kernel_filename; |
3ef96221 | 2174 | const char *initrd_filename = machine->initrd_filename; |
8c9f64df | 2175 | PCIHostState *phb; |
9fdf0c29 | 2176 | int i; |
890c2b77 AK |
2177 | MemoryRegion *sysmem = get_system_memory(); |
2178 | MemoryRegion *ram = g_new(MemoryRegion, 1); | |
658fa66b AK |
2179 | MemoryRegion *rma_region; |
2180 | void *rma = NULL; | |
a8170e5e | 2181 | hwaddr rma_alloc_size; |
b082d65a | 2182 | hwaddr node0_size = spapr_node0_size(); |
b7d1f77a | 2183 | long load_limit, fw_size; |
39ac8455 | 2184 | char *filename; |
30f4b05b | 2185 | Error *resize_hpt_err = NULL; |
9fdf0c29 | 2186 | |
226419d6 | 2187 | msi_nonbroken = true; |
0ee2c058 | 2188 | |
d43b45e2 | 2189 | QLIST_INIT(&spapr->phbs); |
0cffce56 | 2190 | QTAILQ_INIT(&spapr->pending_dimm_unplugs); |
d43b45e2 | 2191 | |
30f4b05b DG |
2192 | /* Check HPT resizing availability */ |
2193 | kvmppc_check_papr_resize_hpt(&resize_hpt_err); | |
2194 | if (spapr->resize_hpt == SPAPR_RESIZE_HPT_DEFAULT) { | |
2195 | /* | |
2196 | * If the user explicitly requested a mode we should either | |
2197 | * supply it, or fail completely (which we do below). But if | |
2198 | * it's not set explicitly, we reset our mode to something | |
2199 | * that works | |
2200 | */ | |
2201 | if (resize_hpt_err) { | |
2202 | spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED; | |
2203 | error_free(resize_hpt_err); | |
2204 | resize_hpt_err = NULL; | |
2205 | } else { | |
2206 | spapr->resize_hpt = smc->resize_hpt_default; | |
2207 | } | |
2208 | } | |
2209 | ||
2210 | assert(spapr->resize_hpt != SPAPR_RESIZE_HPT_DEFAULT); | |
2211 | ||
2212 | if ((spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) && resize_hpt_err) { | |
2213 | /* | |
2214 | * User requested HPT resize, but this host can't supply it. Bail out | |
2215 | */ | |
2216 | error_report_err(resize_hpt_err); | |
2217 | exit(1); | |
2218 | } | |
2219 | ||
354ac20a | 2220 | /* Allocate RMA if necessary */ |
658fa66b | 2221 | rma_alloc_size = kvmppc_alloc_rma(&rma); |
354ac20a DG |
2222 | |
2223 | if (rma_alloc_size == -1) { | |
730fce59 | 2224 | error_report("Unable to create RMA"); |
354ac20a DG |
2225 | exit(1); |
2226 | } | |
7f763a5d | 2227 | |
c4177479 | 2228 | if (rma_alloc_size && (rma_alloc_size < node0_size)) { |
7f763a5d | 2229 | spapr->rma_size = rma_alloc_size; |
354ac20a | 2230 | } else { |
c4177479 | 2231 | spapr->rma_size = node0_size; |
7f763a5d DG |
2232 | |
2233 | /* With KVM, we don't actually know whether KVM supports an | |
2234 | * unbounded RMA (PR KVM) or is limited by the hash table size | |
2235 | * (HV KVM using VRMA), so we always assume the latter | |
2236 | * | |
2237 | * In that case, we also limit the initial allocations for RTAS | |
2238 | * etc... to 256M since we have no way to know what the VRMA size | |
2239 | * is going to be as it depends on the size of the hash table | |
2240 | * isn't determined yet. | |
2241 | */ | |
2242 | if (kvm_enabled()) { | |
2243 | spapr->vrma_adjust = 1; | |
2244 | spapr->rma_size = MIN(spapr->rma_size, 0x10000000); | |
2245 | } | |
912acdf4 BH |
2246 | |
2247 | /* Actually we don't support unbounded RMA anymore since we | |
2248 | * added proper emulation of HV mode. The max we can get is | |
2249 | * 16G which also happens to be what we configure for PAPR | |
2250 | * mode so make sure we don't do anything bigger than that | |
2251 | */ | |
2252 | spapr->rma_size = MIN(spapr->rma_size, 0x400000000ull); | |
354ac20a DG |
2253 | } |
2254 | ||
c4177479 | 2255 | if (spapr->rma_size > node0_size) { |
d54e4d76 DG |
2256 | error_report("Numa node 0 has to span the RMA (%#08"HWADDR_PRIx")", |
2257 | spapr->rma_size); | |
c4177479 AK |
2258 | exit(1); |
2259 | } | |
2260 | ||
b7d1f77a BH |
2261 | /* Setup a load limit for the ramdisk leaving room for SLOF and FDT */ |
2262 | load_limit = MIN(spapr->rma_size, RTAS_MAX_ADDR) - FW_OVERHEAD; | |
9fdf0c29 | 2263 | |
7b565160 | 2264 | /* Set up Interrupt Controller before we create the VCPUs */ |
71cd4dac | 2265 | xics_system_init(machine, XICS_IRQS_SPAPR, &error_fatal); |
7b565160 | 2266 | |
facdb8b6 MR |
2267 | /* Set up containers for ibm,client-set-architecture negotiated options */ |
2268 | spapr->ov5 = spapr_ovec_new(); | |
2269 | spapr->ov5_cas = spapr_ovec_new(); | |
2270 | ||
224245bf | 2271 | if (smc->dr_lmb_enabled) { |
facdb8b6 | 2272 | spapr_ovec_set(spapr->ov5, OV5_DRCONF_MEMORY); |
7c150d6f | 2273 | spapr_validate_node_memory(machine, &error_fatal); |
224245bf DG |
2274 | } |
2275 | ||
417ece33 | 2276 | spapr_ovec_set(spapr->ov5, OV5_FORM1_AFFINITY); |
545d6e2b SJS |
2277 | if (!kvm_enabled() || kvmppc_has_cap_mmu_radix()) { |
2278 | /* KVM and TCG always allow GTSE with radix... */ | |
9fb4541f SB |
2279 | spapr_ovec_set(spapr->ov5, OV5_MMU_RADIX_GTSE); |
2280 | } | |
2281 | /* ... but not with hash (currently). */ | |
417ece33 | 2282 | |
ffbb1705 MR |
2283 | /* advertise support for dedicated HP event source to guests */ |
2284 | if (spapr->use_hotplug_event_source) { | |
2285 | spapr_ovec_set(spapr->ov5, OV5_HP_EVT); | |
2286 | } | |
2287 | ||
2772cf6b DG |
2288 | /* advertise support for HPT resizing */ |
2289 | if (spapr->resize_hpt != SPAPR_RESIZE_HPT_DISABLED) { | |
2290 | spapr_ovec_set(spapr->ov5, OV5_HPT_RESIZE); | |
2291 | } | |
2292 | ||
9fdf0c29 | 2293 | /* init CPUs */ |
19fb2c36 | 2294 | if (machine->cpu_model == NULL) { |
3daa4a9f | 2295 | machine->cpu_model = kvm_enabled() ? "host" : smc->tcg_default_cpu; |
9fdf0c29 | 2296 | } |
94a94e4c | 2297 | |
7843c0d6 | 2298 | spapr_cpu_parse_features(spapr); |
e703d2f7 | 2299 | |
0c86d0fd | 2300 | spapr_init_cpus(spapr); |
9fdf0c29 | 2301 | |
026bfd89 DG |
2302 | if (kvm_enabled()) { |
2303 | /* Enable H_LOGICAL_CI_* so SLOF can talk to in-kernel devices */ | |
2304 | kvmppc_enable_logical_ci_hcalls(); | |
ef9971dd | 2305 | kvmppc_enable_set_mode_hcall(); |
5145ad4f NW |
2306 | |
2307 | /* H_CLEAR_MOD/_REF are mandatory in PAPR, but off by default */ | |
2308 | kvmppc_enable_clear_ref_mod_hcalls(); | |
026bfd89 DG |
2309 | } |
2310 | ||
9fdf0c29 | 2311 | /* allocate RAM */ |
f92f5da1 | 2312 | memory_region_allocate_system_memory(ram, NULL, "ppc_spapr.ram", |
fb164994 | 2313 | machine->ram_size); |
f92f5da1 | 2314 | memory_region_add_subregion(sysmem, 0, ram); |
9fdf0c29 | 2315 | |
658fa66b AK |
2316 | if (rma_alloc_size && rma) { |
2317 | rma_region = g_new(MemoryRegion, 1); | |
2318 | memory_region_init_ram_ptr(rma_region, NULL, "ppc_spapr.rma", | |
2319 | rma_alloc_size, rma); | |
2320 | vmstate_register_ram_global(rma_region); | |
2321 | memory_region_add_subregion(sysmem, 0, rma_region); | |
2322 | } | |
2323 | ||
4a1c9cf0 BR |
2324 | /* initialize hotplug memory address space */ |
2325 | if (machine->ram_size < machine->maxram_size) { | |
2326 | ram_addr_t hotplug_mem_size = machine->maxram_size - machine->ram_size; | |
71c9a3dd BR |
2327 | /* |
2328 | * Limit the number of hotpluggable memory slots to half the number | |
2329 | * slots that KVM supports, leaving the other half for PCI and other | |
2330 | * devices. However ensure that number of slots doesn't drop below 32. | |
2331 | */ | |
2332 | int max_memslots = kvm_enabled() ? kvm_get_max_memslots() / 2 : | |
2333 | SPAPR_MAX_RAM_SLOTS; | |
4a1c9cf0 | 2334 | |
71c9a3dd BR |
2335 | if (max_memslots < SPAPR_MAX_RAM_SLOTS) { |
2336 | max_memslots = SPAPR_MAX_RAM_SLOTS; | |
2337 | } | |
2338 | if (machine->ram_slots > max_memslots) { | |
d54e4d76 DG |
2339 | error_report("Specified number of memory slots %" |
2340 | PRIu64" exceeds max supported %d", | |
71c9a3dd | 2341 | machine->ram_slots, max_memslots); |
d54e4d76 | 2342 | exit(1); |
4a1c9cf0 BR |
2343 | } |
2344 | ||
2345 | spapr->hotplug_memory.base = ROUND_UP(machine->ram_size, | |
2346 | SPAPR_HOTPLUG_MEM_ALIGN); | |
2347 | memory_region_init(&spapr->hotplug_memory.mr, OBJECT(spapr), | |
2348 | "hotplug-memory", hotplug_mem_size); | |
2349 | memory_region_add_subregion(sysmem, spapr->hotplug_memory.base, | |
2350 | &spapr->hotplug_memory.mr); | |
2351 | } | |
2352 | ||
224245bf DG |
2353 | if (smc->dr_lmb_enabled) { |
2354 | spapr_create_lmb_dr_connectors(spapr); | |
2355 | } | |
2356 | ||
39ac8455 | 2357 | filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin"); |
4c56440d | 2358 | if (!filename) { |
730fce59 | 2359 | error_report("Could not find LPAR rtas '%s'", "spapr-rtas.bin"); |
4c56440d SW |
2360 | exit(1); |
2361 | } | |
b7d1f77a | 2362 | spapr->rtas_size = get_image_size(filename); |
8afc22a2 ZJ |
2363 | if (spapr->rtas_size < 0) { |
2364 | error_report("Could not get size of LPAR rtas '%s'", filename); | |
2365 | exit(1); | |
2366 | } | |
b7d1f77a BH |
2367 | spapr->rtas_blob = g_malloc(spapr->rtas_size); |
2368 | if (load_image_size(filename, spapr->rtas_blob, spapr->rtas_size) < 0) { | |
730fce59 | 2369 | error_report("Could not load LPAR rtas '%s'", filename); |
39ac8455 DG |
2370 | exit(1); |
2371 | } | |
4d8d5467 | 2372 | if (spapr->rtas_size > RTAS_MAX_SIZE) { |
730fce59 TH |
2373 | error_report("RTAS too big ! 0x%zx bytes (max is 0x%x)", |
2374 | (size_t)spapr->rtas_size, RTAS_MAX_SIZE); | |
4d8d5467 BH |
2375 | exit(1); |
2376 | } | |
7267c094 | 2377 | g_free(filename); |
39ac8455 | 2378 | |
ffbb1705 | 2379 | /* Set up RTAS event infrastructure */ |
74d042e5 DG |
2380 | spapr_events_init(spapr); |
2381 | ||
12f42174 | 2382 | /* Set up the RTC RTAS interfaces */ |
28df36a1 | 2383 | spapr_rtc_create(spapr); |
12f42174 | 2384 | |
b5cec4c5 | 2385 | /* Set up VIO bus */ |
4040ab72 DG |
2386 | spapr->vio_bus = spapr_vio_bus_init(); |
2387 | ||
277f9acf | 2388 | for (i = 0; i < MAX_SERIAL_PORTS; i++) { |
4040ab72 | 2389 | if (serial_hds[i]) { |
d601fac4 | 2390 | spapr_vty_create(spapr->vio_bus, serial_hds[i]); |
4040ab72 DG |
2391 | } |
2392 | } | |
9fdf0c29 | 2393 | |
639e8102 DG |
2394 | /* We always have at least the nvram device on VIO */ |
2395 | spapr_create_nvram(spapr); | |
2396 | ||
3384f95c | 2397 | /* Set up PCI */ |
fa28f71b AK |
2398 | spapr_pci_rtas_init(); |
2399 | ||
89dfd6e1 | 2400 | phb = spapr_create_phb(spapr, 0); |
3384f95c | 2401 | |
277f9acf | 2402 | for (i = 0; i < nb_nics; i++) { |
8d90ad90 DG |
2403 | NICInfo *nd = &nd_table[i]; |
2404 | ||
2405 | if (!nd->model) { | |
7267c094 | 2406 | nd->model = g_strdup("ibmveth"); |
8d90ad90 DG |
2407 | } |
2408 | ||
2409 | if (strcmp(nd->model, "ibmveth") == 0) { | |
d601fac4 | 2410 | spapr_vlan_create(spapr->vio_bus, nd); |
8d90ad90 | 2411 | } else { |
29b358f9 | 2412 | pci_nic_init_nofail(&nd_table[i], phb->bus, nd->model, NULL); |
8d90ad90 DG |
2413 | } |
2414 | } | |
2415 | ||
6e270446 | 2416 | for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) { |
d601fac4 | 2417 | spapr_vscsi_create(spapr->vio_bus); |
6e270446 BH |
2418 | } |
2419 | ||
f28359d8 | 2420 | /* Graphics */ |
14c6a894 | 2421 | if (spapr_vga_init(phb->bus, &error_fatal)) { |
3fc5acde | 2422 | spapr->has_graphics = true; |
c6e76503 | 2423 | machine->usb |= defaults_enabled() && !machine->usb_disabled; |
f28359d8 LZ |
2424 | } |
2425 | ||
4ee9ced9 | 2426 | if (machine->usb) { |
57040d45 TH |
2427 | if (smc->use_ohci_by_default) { |
2428 | pci_create_simple(phb->bus, -1, "pci-ohci"); | |
2429 | } else { | |
2430 | pci_create_simple(phb->bus, -1, "nec-usb-xhci"); | |
2431 | } | |
c86580b8 | 2432 | |
35139a59 | 2433 | if (spapr->has_graphics) { |
c86580b8 MA |
2434 | USBBus *usb_bus = usb_bus_find(-1); |
2435 | ||
2436 | usb_create_simple(usb_bus, "usb-kbd"); | |
2437 | usb_create_simple(usb_bus, "usb-mouse"); | |
35139a59 DG |
2438 | } |
2439 | } | |
2440 | ||
7f763a5d | 2441 | if (spapr->rma_size < (MIN_RMA_SLOF << 20)) { |
d54e4d76 DG |
2442 | error_report( |
2443 | "pSeries SLOF firmware requires >= %ldM guest RMA (Real Mode Area memory)", | |
2444 | MIN_RMA_SLOF); | |
4d8d5467 BH |
2445 | exit(1); |
2446 | } | |
2447 | ||
9fdf0c29 DG |
2448 | if (kernel_filename) { |
2449 | uint64_t lowaddr = 0; | |
2450 | ||
a19f7fb0 DG |
2451 | spapr->kernel_size = load_elf(kernel_filename, translate_kernel_address, |
2452 | NULL, NULL, &lowaddr, NULL, 1, | |
2453 | PPC_ELF_MACHINE, 0, 0); | |
2454 | if (spapr->kernel_size == ELF_LOAD_WRONG_ENDIAN) { | |
2455 | spapr->kernel_size = load_elf(kernel_filename, | |
2456 | translate_kernel_address, NULL, NULL, | |
2457 | &lowaddr, NULL, 0, PPC_ELF_MACHINE, | |
2458 | 0, 0); | |
2459 | spapr->kernel_le = spapr->kernel_size > 0; | |
16457e7f | 2460 | } |
a19f7fb0 DG |
2461 | if (spapr->kernel_size < 0) { |
2462 | error_report("error loading %s: %s", kernel_filename, | |
2463 | load_elf_strerror(spapr->kernel_size)); | |
9fdf0c29 DG |
2464 | exit(1); |
2465 | } | |
2466 | ||
2467 | /* load initrd */ | |
2468 | if (initrd_filename) { | |
4d8d5467 BH |
2469 | /* Try to locate the initrd in the gap between the kernel |
2470 | * and the firmware. Add a bit of space just in case | |
2471 | */ | |
a19f7fb0 DG |
2472 | spapr->initrd_base = (KERNEL_LOAD_ADDR + spapr->kernel_size |
2473 | + 0x1ffff) & ~0xffff; | |
2474 | spapr->initrd_size = load_image_targphys(initrd_filename, | |
2475 | spapr->initrd_base, | |
2476 | load_limit | |
2477 | - spapr->initrd_base); | |
2478 | if (spapr->initrd_size < 0) { | |
d54e4d76 DG |
2479 | error_report("could not load initial ram disk '%s'", |
2480 | initrd_filename); | |
9fdf0c29 DG |
2481 | exit(1); |
2482 | } | |
9fdf0c29 | 2483 | } |
4d8d5467 | 2484 | } |
a3467baa | 2485 | |
8e7ea787 AF |
2486 | if (bios_name == NULL) { |
2487 | bios_name = FW_FILE_NAME; | |
2488 | } | |
2489 | filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name); | |
4c56440d | 2490 | if (!filename) { |
68fea5a0 | 2491 | error_report("Could not find LPAR firmware '%s'", bios_name); |
4c56440d SW |
2492 | exit(1); |
2493 | } | |
4d8d5467 | 2494 | fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE); |
68fea5a0 TH |
2495 | if (fw_size <= 0) { |
2496 | error_report("Could not load LPAR firmware '%s'", filename); | |
4d8d5467 BH |
2497 | exit(1); |
2498 | } | |
2499 | g_free(filename); | |
4d8d5467 | 2500 | |
28e02042 DG |
2501 | /* FIXME: Should register things through the MachineState's qdev |
2502 | * interface, this is a legacy from the sPAPREnvironment structure | |
2503 | * which predated MachineState but had a similar function */ | |
4be21d56 DG |
2504 | vmstate_register(NULL, 0, &vmstate_spapr, spapr); |
2505 | register_savevm_live(NULL, "spapr/htab", -1, 1, | |
2506 | &savevm_htab_handlers, spapr); | |
2507 | ||
5b2128d2 | 2508 | qemu_register_boot_set(spapr_boot_set, spapr); |
42043e4f | 2509 | |
42043e4f | 2510 | if (kvm_enabled()) { |
3dc410ae | 2511 | /* to stop and start vmclock */ |
42043e4f LV |
2512 | qemu_add_vm_change_state_handler(cpu_ppc_clock_vm_state_change, |
2513 | &spapr->tb); | |
3dc410ae AK |
2514 | |
2515 | kvmppc_spapr_enable_inkernel_multitce(); | |
42043e4f | 2516 | } |
9fdf0c29 DG |
2517 | } |
2518 | ||
135a129a AK |
2519 | static int spapr_kvm_type(const char *vm_type) |
2520 | { | |
2521 | if (!vm_type) { | |
2522 | return 0; | |
2523 | } | |
2524 | ||
2525 | if (!strcmp(vm_type, "HV")) { | |
2526 | return 1; | |
2527 | } | |
2528 | ||
2529 | if (!strcmp(vm_type, "PR")) { | |
2530 | return 2; | |
2531 | } | |
2532 | ||
2533 | error_report("Unknown kvm-type specified '%s'", vm_type); | |
2534 | exit(1); | |
2535 | } | |
2536 | ||
71461b0f | 2537 | /* |
627b84f4 | 2538 | * Implementation of an interface to adjust firmware path |
71461b0f AK |
2539 | * for the bootindex property handling. |
2540 | */ | |
2541 | static char *spapr_get_fw_dev_path(FWPathProvider *p, BusState *bus, | |
2542 | DeviceState *dev) | |
2543 | { | |
2544 | #define CAST(type, obj, name) \ | |
2545 | ((type *)object_dynamic_cast(OBJECT(obj), (name))) | |
2546 | SCSIDevice *d = CAST(SCSIDevice, dev, TYPE_SCSI_DEVICE); | |
2547 | sPAPRPHBState *phb = CAST(sPAPRPHBState, dev, TYPE_SPAPR_PCI_HOST_BRIDGE); | |
c4e13492 | 2548 | VHostSCSICommon *vsc = CAST(VHostSCSICommon, dev, TYPE_VHOST_SCSI_COMMON); |
71461b0f AK |
2549 | |
2550 | if (d) { | |
2551 | void *spapr = CAST(void, bus->parent, "spapr-vscsi"); | |
2552 | VirtIOSCSI *virtio = CAST(VirtIOSCSI, bus->parent, TYPE_VIRTIO_SCSI); | |
2553 | USBDevice *usb = CAST(USBDevice, bus->parent, TYPE_USB_DEVICE); | |
2554 | ||
2555 | if (spapr) { | |
2556 | /* | |
2557 | * Replace "channel@0/disk@0,0" with "disk@8000000000000000": | |
2558 | * We use SRP luns of the form 8000 | (bus << 8) | (id << 5) | lun | |
2559 | * in the top 16 bits of the 64-bit LUN | |
2560 | */ | |
2561 | unsigned id = 0x8000 | (d->id << 8) | d->lun; | |
2562 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2563 | (uint64_t)id << 48); | |
2564 | } else if (virtio) { | |
2565 | /* | |
2566 | * We use SRP luns of the form 01000000 | (target << 8) | lun | |
2567 | * in the top 32 bits of the 64-bit LUN | |
2568 | * Note: the quote above is from SLOF and it is wrong, | |
2569 | * the actual binding is: | |
2570 | * swap 0100 or 10 << or 20 << ( target lun-id -- srplun ) | |
2571 | */ | |
2572 | unsigned id = 0x1000000 | (d->id << 16) | d->lun; | |
2573 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2574 | (uint64_t)id << 32); | |
2575 | } else if (usb) { | |
2576 | /* | |
2577 | * We use SRP luns of the form 01000000 | (usb-port << 16) | lun | |
2578 | * in the top 32 bits of the 64-bit LUN | |
2579 | */ | |
2580 | unsigned usb_port = atoi(usb->port->path); | |
2581 | unsigned id = 0x1000000 | (usb_port << 16) | d->lun; | |
2582 | return g_strdup_printf("%s@%"PRIX64, qdev_fw_name(dev), | |
2583 | (uint64_t)id << 32); | |
2584 | } | |
2585 | } | |
2586 | ||
b99260eb TH |
2587 | /* |
2588 | * SLOF probes the USB devices, and if it recognizes that the device is a | |
2589 | * storage device, it changes its name to "storage" instead of "usb-host", | |
2590 | * and additionally adds a child node for the SCSI LUN, so the correct | |
2591 | * boot path in SLOF is something like .../storage@1/disk@xxx" instead. | |
2592 | */ | |
2593 | if (strcmp("usb-host", qdev_fw_name(dev)) == 0) { | |
2594 | USBDevice *usbdev = CAST(USBDevice, dev, TYPE_USB_DEVICE); | |
2595 | if (usb_host_dev_is_scsi_storage(usbdev)) { | |
2596 | return g_strdup_printf("storage@%s/disk", usbdev->port->path); | |
2597 | } | |
2598 | } | |
2599 | ||
71461b0f AK |
2600 | if (phb) { |
2601 | /* Replace "pci" with "pci@800000020000000" */ | |
2602 | return g_strdup_printf("pci@%"PRIX64, phb->buid); | |
2603 | } | |
2604 | ||
c4e13492 FF |
2605 | if (vsc) { |
2606 | /* Same logic as virtio above */ | |
2607 | unsigned id = 0x1000000 | (vsc->target << 16) | vsc->lun; | |
2608 | return g_strdup_printf("disk@%"PRIX64, (uint64_t)id << 32); | |
2609 | } | |
2610 | ||
4871dd4c TH |
2611 | if (g_str_equal("pci-bridge", qdev_fw_name(dev))) { |
2612 | /* SLOF uses "pci" instead of "pci-bridge" for PCI bridges */ | |
2613 | PCIDevice *pcidev = CAST(PCIDevice, dev, TYPE_PCI_DEVICE); | |
2614 | return g_strdup_printf("pci@%x", PCI_SLOT(pcidev->devfn)); | |
2615 | } | |
2616 | ||
71461b0f AK |
2617 | return NULL; |
2618 | } | |
2619 | ||
23825581 EH |
2620 | static char *spapr_get_kvm_type(Object *obj, Error **errp) |
2621 | { | |
28e02042 | 2622 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
23825581 | 2623 | |
28e02042 | 2624 | return g_strdup(spapr->kvm_type); |
23825581 EH |
2625 | } |
2626 | ||
2627 | static void spapr_set_kvm_type(Object *obj, const char *value, Error **errp) | |
2628 | { | |
28e02042 | 2629 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
23825581 | 2630 | |
28e02042 DG |
2631 | g_free(spapr->kvm_type); |
2632 | spapr->kvm_type = g_strdup(value); | |
23825581 EH |
2633 | } |
2634 | ||
f6229214 MR |
2635 | static bool spapr_get_modern_hotplug_events(Object *obj, Error **errp) |
2636 | { | |
2637 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2638 | ||
2639 | return spapr->use_hotplug_event_source; | |
2640 | } | |
2641 | ||
2642 | static void spapr_set_modern_hotplug_events(Object *obj, bool value, | |
2643 | Error **errp) | |
2644 | { | |
2645 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2646 | ||
2647 | spapr->use_hotplug_event_source = value; | |
2648 | } | |
2649 | ||
30f4b05b DG |
2650 | static char *spapr_get_resize_hpt(Object *obj, Error **errp) |
2651 | { | |
2652 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2653 | ||
2654 | switch (spapr->resize_hpt) { | |
2655 | case SPAPR_RESIZE_HPT_DEFAULT: | |
2656 | return g_strdup("default"); | |
2657 | case SPAPR_RESIZE_HPT_DISABLED: | |
2658 | return g_strdup("disabled"); | |
2659 | case SPAPR_RESIZE_HPT_ENABLED: | |
2660 | return g_strdup("enabled"); | |
2661 | case SPAPR_RESIZE_HPT_REQUIRED: | |
2662 | return g_strdup("required"); | |
2663 | } | |
2664 | g_assert_not_reached(); | |
2665 | } | |
2666 | ||
2667 | static void spapr_set_resize_hpt(Object *obj, const char *value, Error **errp) | |
2668 | { | |
2669 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2670 | ||
2671 | if (strcmp(value, "default") == 0) { | |
2672 | spapr->resize_hpt = SPAPR_RESIZE_HPT_DEFAULT; | |
2673 | } else if (strcmp(value, "disabled") == 0) { | |
2674 | spapr->resize_hpt = SPAPR_RESIZE_HPT_DISABLED; | |
2675 | } else if (strcmp(value, "enabled") == 0) { | |
2676 | spapr->resize_hpt = SPAPR_RESIZE_HPT_ENABLED; | |
2677 | } else if (strcmp(value, "required") == 0) { | |
2678 | spapr->resize_hpt = SPAPR_RESIZE_HPT_REQUIRED; | |
2679 | } else { | |
2680 | error_setg(errp, "Bad value for \"resize-hpt\" property"); | |
2681 | } | |
2682 | } | |
2683 | ||
23825581 EH |
2684 | static void spapr_machine_initfn(Object *obj) |
2685 | { | |
715c5407 DG |
2686 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); |
2687 | ||
2688 | spapr->htab_fd = -1; | |
f6229214 | 2689 | spapr->use_hotplug_event_source = true; |
23825581 EH |
2690 | object_property_add_str(obj, "kvm-type", |
2691 | spapr_get_kvm_type, spapr_set_kvm_type, NULL); | |
49d2e648 MA |
2692 | object_property_set_description(obj, "kvm-type", |
2693 | "Specifies the KVM virtualization mode (HV, PR)", | |
2694 | NULL); | |
f6229214 MR |
2695 | object_property_add_bool(obj, "modern-hotplug-events", |
2696 | spapr_get_modern_hotplug_events, | |
2697 | spapr_set_modern_hotplug_events, | |
2698 | NULL); | |
2699 | object_property_set_description(obj, "modern-hotplug-events", | |
2700 | "Use dedicated hotplug event mechanism in" | |
2701 | " place of standard EPOW events when possible" | |
2702 | " (required for memory hot-unplug support)", | |
2703 | NULL); | |
7843c0d6 DG |
2704 | |
2705 | ppc_compat_add_property(obj, "max-cpu-compat", &spapr->max_compat_pvr, | |
2706 | "Maximum permitted CPU compatibility mode", | |
2707 | &error_fatal); | |
30f4b05b DG |
2708 | |
2709 | object_property_add_str(obj, "resize-hpt", | |
2710 | spapr_get_resize_hpt, spapr_set_resize_hpt, NULL); | |
2711 | object_property_set_description(obj, "resize-hpt", | |
2712 | "Resizing of the Hash Page Table (enabled, disabled, required)", | |
2713 | NULL); | |
23825581 EH |
2714 | } |
2715 | ||
87bbdd9c DG |
2716 | static void spapr_machine_finalizefn(Object *obj) |
2717 | { | |
2718 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
2719 | ||
2720 | g_free(spapr->kvm_type); | |
2721 | } | |
2722 | ||
1c7ad77e | 2723 | void spapr_do_system_reset_on_cpu(CPUState *cs, run_on_cpu_data arg) |
34316482 | 2724 | { |
34316482 AK |
2725 | cpu_synchronize_state(cs); |
2726 | ppc_cpu_do_system_reset(cs); | |
2727 | } | |
2728 | ||
2729 | static void spapr_nmi(NMIState *n, int cpu_index, Error **errp) | |
2730 | { | |
2731 | CPUState *cs; | |
2732 | ||
2733 | CPU_FOREACH(cs) { | |
1c7ad77e | 2734 | async_run_on_cpu(cs, spapr_do_system_reset_on_cpu, RUN_ON_CPU_NULL); |
34316482 AK |
2735 | } |
2736 | } | |
2737 | ||
79b78a6b MR |
2738 | static void spapr_add_lmbs(DeviceState *dev, uint64_t addr_start, uint64_t size, |
2739 | uint32_t node, bool dedicated_hp_event_source, | |
2740 | Error **errp) | |
c20d332a BR |
2741 | { |
2742 | sPAPRDRConnector *drc; | |
c20d332a BR |
2743 | uint32_t nr_lmbs = size/SPAPR_MEMORY_BLOCK_SIZE; |
2744 | int i, fdt_offset, fdt_size; | |
2745 | void *fdt; | |
79b78a6b | 2746 | uint64_t addr = addr_start; |
94fd9cba | 2747 | bool hotplugged = spapr_drc_hotplugged(dev); |
160bb678 | 2748 | Error *local_err = NULL; |
c20d332a | 2749 | |
c20d332a | 2750 | for (i = 0; i < nr_lmbs; i++) { |
fbf55397 DG |
2751 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2752 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
c20d332a BR |
2753 | g_assert(drc); |
2754 | ||
2755 | fdt = create_device_tree(&fdt_size); | |
2756 | fdt_offset = spapr_populate_memory_node(fdt, node, addr, | |
2757 | SPAPR_MEMORY_BLOCK_SIZE); | |
2758 | ||
160bb678 GK |
2759 | spapr_drc_attach(drc, dev, fdt, fdt_offset, &local_err); |
2760 | if (local_err) { | |
2761 | while (addr > addr_start) { | |
2762 | addr -= SPAPR_MEMORY_BLOCK_SIZE; | |
2763 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, | |
2764 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
a8dc47fd | 2765 | spapr_drc_detach(drc); |
160bb678 GK |
2766 | } |
2767 | g_free(fdt); | |
2768 | error_propagate(errp, local_err); | |
2769 | return; | |
2770 | } | |
94fd9cba LV |
2771 | if (!hotplugged) { |
2772 | spapr_drc_reset(drc); | |
2773 | } | |
c20d332a BR |
2774 | addr += SPAPR_MEMORY_BLOCK_SIZE; |
2775 | } | |
5dd5238c JD |
2776 | /* send hotplug notification to the |
2777 | * guest only in case of hotplugged memory | |
2778 | */ | |
94fd9cba | 2779 | if (hotplugged) { |
79b78a6b | 2780 | if (dedicated_hp_event_source) { |
fbf55397 DG |
2781 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2782 | addr_start / SPAPR_MEMORY_BLOCK_SIZE); | |
79b78a6b MR |
2783 | spapr_hotplug_req_add_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, |
2784 | nr_lmbs, | |
0b55aa91 | 2785 | spapr_drc_index(drc)); |
79b78a6b MR |
2786 | } else { |
2787 | spapr_hotplug_req_add_by_count(SPAPR_DR_CONNECTOR_TYPE_LMB, | |
2788 | nr_lmbs); | |
2789 | } | |
5dd5238c | 2790 | } |
c20d332a BR |
2791 | } |
2792 | ||
2793 | static void spapr_memory_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
2794 | uint32_t node, Error **errp) | |
2795 | { | |
2796 | Error *local_err = NULL; | |
2797 | sPAPRMachineState *ms = SPAPR_MACHINE(hotplug_dev); | |
2798 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2799 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
04790978 TH |
2800 | MemoryRegion *mr; |
2801 | uint64_t align, size, addr; | |
2802 | ||
2803 | mr = ddc->get_memory_region(dimm, &local_err); | |
2804 | if (local_err) { | |
2805 | goto out; | |
2806 | } | |
2807 | align = memory_region_get_alignment(mr); | |
2808 | size = memory_region_size(mr); | |
df587133 | 2809 | |
d6a9b0b8 | 2810 | pc_dimm_memory_plug(dev, &ms->hotplug_memory, mr, align, &local_err); |
c20d332a BR |
2811 | if (local_err) { |
2812 | goto out; | |
2813 | } | |
2814 | ||
9ed442b8 MAL |
2815 | addr = object_property_get_uint(OBJECT(dimm), |
2816 | PC_DIMM_ADDR_PROP, &local_err); | |
c20d332a | 2817 | if (local_err) { |
160bb678 | 2818 | goto out_unplug; |
c20d332a BR |
2819 | } |
2820 | ||
79b78a6b MR |
2821 | spapr_add_lmbs(dev, addr, size, node, |
2822 | spapr_ovec_test(ms->ov5_cas, OV5_HP_EVT), | |
160bb678 GK |
2823 | &local_err); |
2824 | if (local_err) { | |
2825 | goto out_unplug; | |
2826 | } | |
2827 | ||
2828 | return; | |
c20d332a | 2829 | |
160bb678 GK |
2830 | out_unplug: |
2831 | pc_dimm_memory_unplug(dev, &ms->hotplug_memory, mr); | |
c20d332a BR |
2832 | out: |
2833 | error_propagate(errp, local_err); | |
2834 | } | |
2835 | ||
c871bc70 LV |
2836 | static void spapr_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev, |
2837 | Error **errp) | |
2838 | { | |
2839 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2840 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
04790978 TH |
2841 | MemoryRegion *mr; |
2842 | uint64_t size; | |
c871bc70 LV |
2843 | char *mem_dev; |
2844 | ||
04790978 TH |
2845 | mr = ddc->get_memory_region(dimm, errp); |
2846 | if (!mr) { | |
2847 | return; | |
2848 | } | |
2849 | size = memory_region_size(mr); | |
2850 | ||
c871bc70 LV |
2851 | if (size % SPAPR_MEMORY_BLOCK_SIZE) { |
2852 | error_setg(errp, "Hotplugged memory size must be a multiple of " | |
2853 | "%lld MB", SPAPR_MEMORY_BLOCK_SIZE / M_BYTE); | |
2854 | return; | |
2855 | } | |
2856 | ||
2857 | mem_dev = object_property_get_str(OBJECT(dimm), PC_DIMM_MEMDEV_PROP, NULL); | |
2858 | if (mem_dev && !kvmppc_is_mem_backend_page_size_ok(mem_dev)) { | |
2859 | error_setg(errp, "Memory backend has bad page size. " | |
2860 | "Use 'memory-backend-file' with correct mem-path."); | |
8a9e0e7b | 2861 | goto out; |
c871bc70 | 2862 | } |
8a9e0e7b GK |
2863 | |
2864 | out: | |
2865 | g_free(mem_dev); | |
c871bc70 LV |
2866 | } |
2867 | ||
0cffce56 DG |
2868 | struct sPAPRDIMMState { |
2869 | PCDIMMDevice *dimm; | |
cf632463 | 2870 | uint32_t nr_lmbs; |
0cffce56 DG |
2871 | QTAILQ_ENTRY(sPAPRDIMMState) next; |
2872 | }; | |
2873 | ||
2874 | static sPAPRDIMMState *spapr_pending_dimm_unplugs_find(sPAPRMachineState *s, | |
2875 | PCDIMMDevice *dimm) | |
2876 | { | |
2877 | sPAPRDIMMState *dimm_state = NULL; | |
2878 | ||
2879 | QTAILQ_FOREACH(dimm_state, &s->pending_dimm_unplugs, next) { | |
2880 | if (dimm_state->dimm == dimm) { | |
2881 | break; | |
2882 | } | |
2883 | } | |
2884 | return dimm_state; | |
2885 | } | |
2886 | ||
8d5981c4 BR |
2887 | static sPAPRDIMMState *spapr_pending_dimm_unplugs_add(sPAPRMachineState *spapr, |
2888 | uint32_t nr_lmbs, | |
2889 | PCDIMMDevice *dimm) | |
0cffce56 | 2890 | { |
8d5981c4 BR |
2891 | sPAPRDIMMState *ds = NULL; |
2892 | ||
2893 | /* | |
2894 | * If this request is for a DIMM whose removal had failed earlier | |
2895 | * (due to guest's refusal to remove the LMBs), we would have this | |
2896 | * dimm already in the pending_dimm_unplugs list. In that | |
2897 | * case don't add again. | |
2898 | */ | |
2899 | ds = spapr_pending_dimm_unplugs_find(spapr, dimm); | |
2900 | if (!ds) { | |
2901 | ds = g_malloc0(sizeof(sPAPRDIMMState)); | |
2902 | ds->nr_lmbs = nr_lmbs; | |
2903 | ds->dimm = dimm; | |
2904 | QTAILQ_INSERT_HEAD(&spapr->pending_dimm_unplugs, ds, next); | |
2905 | } | |
2906 | return ds; | |
0cffce56 DG |
2907 | } |
2908 | ||
2909 | static void spapr_pending_dimm_unplugs_remove(sPAPRMachineState *spapr, | |
2910 | sPAPRDIMMState *dimm_state) | |
2911 | { | |
2912 | QTAILQ_REMOVE(&spapr->pending_dimm_unplugs, dimm_state, next); | |
2913 | g_free(dimm_state); | |
2914 | } | |
cf632463 | 2915 | |
16ee9980 DHB |
2916 | static sPAPRDIMMState *spapr_recover_pending_dimm_state(sPAPRMachineState *ms, |
2917 | PCDIMMDevice *dimm) | |
2918 | { | |
2919 | sPAPRDRConnector *drc; | |
2920 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
04790978 | 2921 | MemoryRegion *mr = ddc->get_memory_region(dimm, &error_abort); |
16ee9980 DHB |
2922 | uint64_t size = memory_region_size(mr); |
2923 | uint32_t nr_lmbs = size / SPAPR_MEMORY_BLOCK_SIZE; | |
2924 | uint32_t avail_lmbs = 0; | |
2925 | uint64_t addr_start, addr; | |
2926 | int i; | |
16ee9980 DHB |
2927 | |
2928 | addr_start = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, | |
2929 | &error_abort); | |
2930 | ||
2931 | addr = addr_start; | |
2932 | for (i = 0; i < nr_lmbs; i++) { | |
fbf55397 DG |
2933 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
2934 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
16ee9980 | 2935 | g_assert(drc); |
454b580a | 2936 | if (drc->dev) { |
16ee9980 DHB |
2937 | avail_lmbs++; |
2938 | } | |
2939 | addr += SPAPR_MEMORY_BLOCK_SIZE; | |
2940 | } | |
2941 | ||
8d5981c4 | 2942 | return spapr_pending_dimm_unplugs_add(ms, avail_lmbs, dimm); |
16ee9980 DHB |
2943 | } |
2944 | ||
31834723 DHB |
2945 | /* Callback to be called during DRC release. */ |
2946 | void spapr_lmb_release(DeviceState *dev) | |
cf632463 | 2947 | { |
765d1bdd DG |
2948 | sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_hotplug_handler(dev)); |
2949 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2950 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
04790978 | 2951 | MemoryRegion *mr = ddc->get_memory_region(dimm, &error_abort); |
0cffce56 | 2952 | sPAPRDIMMState *ds = spapr_pending_dimm_unplugs_find(spapr, PC_DIMM(dev)); |
cf632463 | 2953 | |
16ee9980 DHB |
2954 | /* This information will get lost if a migration occurs |
2955 | * during the unplug process. In this case recover it. */ | |
2956 | if (ds == NULL) { | |
2957 | ds = spapr_recover_pending_dimm_state(spapr, PC_DIMM(dev)); | |
8d5981c4 | 2958 | g_assert(ds); |
454b580a DG |
2959 | /* The DRC being examined by the caller at least must be counted */ |
2960 | g_assert(ds->nr_lmbs); | |
2961 | } | |
2962 | ||
2963 | if (--ds->nr_lmbs) { | |
cf632463 BR |
2964 | return; |
2965 | } | |
2966 | ||
0cffce56 | 2967 | spapr_pending_dimm_unplugs_remove(spapr, ds); |
cf632463 BR |
2968 | |
2969 | /* | |
2970 | * Now that all the LMBs have been removed by the guest, call the | |
2971 | * pc-dimm unplug handler to cleanup up the pc-dimm device. | |
2972 | */ | |
765d1bdd | 2973 | pc_dimm_memory_unplug(dev, &spapr->hotplug_memory, mr); |
cf632463 BR |
2974 | object_unparent(OBJECT(dev)); |
2975 | } | |
2976 | ||
2977 | static void spapr_memory_unplug_request(HotplugHandler *hotplug_dev, | |
2978 | DeviceState *dev, Error **errp) | |
2979 | { | |
0cffce56 | 2980 | sPAPRMachineState *spapr = SPAPR_MACHINE(hotplug_dev); |
cf632463 BR |
2981 | Error *local_err = NULL; |
2982 | PCDIMMDevice *dimm = PC_DIMM(dev); | |
2983 | PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); | |
04790978 TH |
2984 | MemoryRegion *mr; |
2985 | uint32_t nr_lmbs; | |
2986 | uint64_t size, addr_start, addr; | |
0cffce56 DG |
2987 | int i; |
2988 | sPAPRDRConnector *drc; | |
04790978 TH |
2989 | |
2990 | mr = ddc->get_memory_region(dimm, &local_err); | |
2991 | if (local_err) { | |
2992 | goto out; | |
2993 | } | |
2994 | size = memory_region_size(mr); | |
2995 | nr_lmbs = size / SPAPR_MEMORY_BLOCK_SIZE; | |
2996 | ||
9ed442b8 | 2997 | addr_start = object_property_get_uint(OBJECT(dimm), PC_DIMM_ADDR_PROP, |
0cffce56 | 2998 | &local_err); |
cf632463 BR |
2999 | if (local_err) { |
3000 | goto out; | |
3001 | } | |
3002 | ||
8d5981c4 | 3003 | spapr_pending_dimm_unplugs_add(spapr, nr_lmbs, dimm); |
0cffce56 DG |
3004 | |
3005 | addr = addr_start; | |
3006 | for (i = 0; i < nr_lmbs; i++) { | |
fbf55397 DG |
3007 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
3008 | addr / SPAPR_MEMORY_BLOCK_SIZE); | |
0cffce56 DG |
3009 | g_assert(drc); |
3010 | ||
a8dc47fd | 3011 | spapr_drc_detach(drc); |
0cffce56 DG |
3012 | addr += SPAPR_MEMORY_BLOCK_SIZE; |
3013 | } | |
3014 | ||
fbf55397 DG |
3015 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_LMB, |
3016 | addr_start / SPAPR_MEMORY_BLOCK_SIZE); | |
0cffce56 | 3017 | spapr_hotplug_req_remove_by_count_indexed(SPAPR_DR_CONNECTOR_TYPE_LMB, |
0b55aa91 | 3018 | nr_lmbs, spapr_drc_index(drc)); |
cf632463 BR |
3019 | out: |
3020 | error_propagate(errp, local_err); | |
3021 | } | |
3022 | ||
04d0ffbd GK |
3023 | static void *spapr_populate_hotplug_cpu_dt(CPUState *cs, int *fdt_offset, |
3024 | sPAPRMachineState *spapr) | |
af81cf32 BR |
3025 | { |
3026 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
3027 | DeviceClass *dc = DEVICE_GET_CLASS(cs); | |
81210c20 | 3028 | int id = ppc_get_vcpu_id(cpu); |
af81cf32 BR |
3029 | void *fdt; |
3030 | int offset, fdt_size; | |
3031 | char *nodename; | |
3032 | ||
3033 | fdt = create_device_tree(&fdt_size); | |
3034 | nodename = g_strdup_printf("%s@%x", dc->fw_name, id); | |
3035 | offset = fdt_add_subnode(fdt, 0, nodename); | |
3036 | ||
3037 | spapr_populate_cpu_dt(cs, fdt, offset, spapr); | |
3038 | g_free(nodename); | |
3039 | ||
3040 | *fdt_offset = offset; | |
3041 | return fdt; | |
3042 | } | |
3043 | ||
765d1bdd DG |
3044 | /* Callback to be called during DRC release. */ |
3045 | void spapr_core_release(DeviceState *dev) | |
ff9006dd | 3046 | { |
765d1bdd | 3047 | MachineState *ms = MACHINE(qdev_get_hotplug_handler(dev)); |
46f7afa3 | 3048 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(ms); |
ff9006dd | 3049 | CPUCore *cc = CPU_CORE(dev); |
535455fd | 3050 | CPUArchId *core_slot = spapr_find_cpu_slot(ms, cc->core_id, NULL); |
ff9006dd | 3051 | |
46f7afa3 GK |
3052 | if (smc->pre_2_10_has_unused_icps) { |
3053 | sPAPRCPUCore *sc = SPAPR_CPU_CORE(OBJECT(dev)); | |
3054 | sPAPRCPUCoreClass *scc = SPAPR_CPU_CORE_GET_CLASS(OBJECT(cc)); | |
3055 | const char *typename = object_class_get_name(scc->cpu_class); | |
3056 | size_t size = object_type_get_instance_size(typename); | |
3057 | int i; | |
3058 | ||
3059 | for (i = 0; i < cc->nr_threads; i++) { | |
3060 | CPUState *cs = CPU(sc->threads + i * size); | |
3061 | ||
3062 | pre_2_10_vmstate_register_dummy_icp(cs->cpu_index); | |
3063 | } | |
3064 | } | |
3065 | ||
07572c06 | 3066 | assert(core_slot); |
535455fd | 3067 | core_slot->cpu = NULL; |
ff9006dd IM |
3068 | object_unparent(OBJECT(dev)); |
3069 | } | |
3070 | ||
115debf2 IM |
3071 | static |
3072 | void spapr_core_unplug_request(HotplugHandler *hotplug_dev, DeviceState *dev, | |
3073 | Error **errp) | |
ff9006dd | 3074 | { |
535455fd IM |
3075 | int index; |
3076 | sPAPRDRConnector *drc; | |
535455fd IM |
3077 | CPUCore *cc = CPU_CORE(dev); |
3078 | int smt = kvmppc_smt_threads(); | |
ff9006dd | 3079 | |
535455fd IM |
3080 | if (!spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index)) { |
3081 | error_setg(errp, "Unable to find CPU core with core-id: %d", | |
3082 | cc->core_id); | |
3083 | return; | |
3084 | } | |
ff9006dd IM |
3085 | if (index == 0) { |
3086 | error_setg(errp, "Boot CPU core may not be unplugged"); | |
3087 | return; | |
3088 | } | |
3089 | ||
fbf55397 | 3090 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index * smt); |
ff9006dd IM |
3091 | g_assert(drc); |
3092 | ||
a8dc47fd | 3093 | spapr_drc_detach(drc); |
ff9006dd IM |
3094 | |
3095 | spapr_hotplug_req_remove_by_index(drc); | |
3096 | } | |
3097 | ||
3098 | static void spapr_core_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
3099 | Error **errp) | |
3100 | { | |
3101 | sPAPRMachineState *spapr = SPAPR_MACHINE(OBJECT(hotplug_dev)); | |
3102 | MachineClass *mc = MACHINE_GET_CLASS(spapr); | |
46f7afa3 | 3103 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
ff9006dd IM |
3104 | sPAPRCPUCore *core = SPAPR_CPU_CORE(OBJECT(dev)); |
3105 | CPUCore *cc = CPU_CORE(dev); | |
3106 | CPUState *cs = CPU(core->threads); | |
3107 | sPAPRDRConnector *drc; | |
3108 | Error *local_err = NULL; | |
ff9006dd | 3109 | int smt = kvmppc_smt_threads(); |
535455fd IM |
3110 | CPUArchId *core_slot; |
3111 | int index; | |
94fd9cba | 3112 | bool hotplugged = spapr_drc_hotplugged(dev); |
ff9006dd | 3113 | |
535455fd IM |
3114 | core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); |
3115 | if (!core_slot) { | |
3116 | error_setg(errp, "Unable to find CPU core with core-id: %d", | |
3117 | cc->core_id); | |
3118 | return; | |
3119 | } | |
fbf55397 | 3120 | drc = spapr_drc_by_id(TYPE_SPAPR_DRC_CPU, index * smt); |
ff9006dd | 3121 | |
c5514d0e | 3122 | g_assert(drc || !mc->has_hotpluggable_cpus); |
ff9006dd | 3123 | |
ff9006dd | 3124 | if (drc) { |
e49c63d5 GK |
3125 | void *fdt; |
3126 | int fdt_offset; | |
3127 | ||
3128 | fdt = spapr_populate_hotplug_cpu_dt(cs, &fdt_offset, spapr); | |
3129 | ||
5c1da812 | 3130 | spapr_drc_attach(drc, dev, fdt, fdt_offset, &local_err); |
ff9006dd IM |
3131 | if (local_err) { |
3132 | g_free(fdt); | |
ff9006dd IM |
3133 | error_propagate(errp, local_err); |
3134 | return; | |
3135 | } | |
ff9006dd | 3136 | |
94fd9cba LV |
3137 | if (hotplugged) { |
3138 | /* | |
3139 | * Send hotplug notification interrupt to the guest only | |
3140 | * in case of hotplugged CPUs. | |
3141 | */ | |
3142 | spapr_hotplug_req_add_by_index(drc); | |
3143 | } else { | |
3144 | spapr_drc_reset(drc); | |
3145 | } | |
ff9006dd | 3146 | } |
94fd9cba | 3147 | |
535455fd | 3148 | core_slot->cpu = OBJECT(dev); |
46f7afa3 GK |
3149 | |
3150 | if (smc->pre_2_10_has_unused_icps) { | |
3151 | sPAPRCPUCoreClass *scc = SPAPR_CPU_CORE_GET_CLASS(OBJECT(cc)); | |
3152 | const char *typename = object_class_get_name(scc->cpu_class); | |
3153 | size_t size = object_type_get_instance_size(typename); | |
3154 | int i; | |
3155 | ||
3156 | for (i = 0; i < cc->nr_threads; i++) { | |
3157 | sPAPRCPUCore *sc = SPAPR_CPU_CORE(dev); | |
3158 | void *obj = sc->threads + i * size; | |
3159 | ||
3160 | cs = CPU(obj); | |
3161 | pre_2_10_vmstate_unregister_dummy_icp(cs->cpu_index); | |
3162 | } | |
3163 | } | |
ff9006dd IM |
3164 | } |
3165 | ||
3166 | static void spapr_core_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev, | |
3167 | Error **errp) | |
3168 | { | |
3169 | MachineState *machine = MACHINE(OBJECT(hotplug_dev)); | |
3170 | MachineClass *mc = MACHINE_GET_CLASS(hotplug_dev); | |
ff9006dd IM |
3171 | Error *local_err = NULL; |
3172 | CPUCore *cc = CPU_CORE(dev); | |
3173 | char *base_core_type = spapr_get_cpu_core_type(machine->cpu_model); | |
3174 | const char *type = object_get_typename(OBJECT(dev)); | |
535455fd IM |
3175 | CPUArchId *core_slot; |
3176 | int index; | |
ff9006dd | 3177 | |
c5514d0e | 3178 | if (dev->hotplugged && !mc->has_hotpluggable_cpus) { |
ff9006dd IM |
3179 | error_setg(&local_err, "CPU hotplug not supported for this machine"); |
3180 | goto out; | |
3181 | } | |
3182 | ||
3183 | if (strcmp(base_core_type, type)) { | |
3184 | error_setg(&local_err, "CPU core type should be %s", base_core_type); | |
3185 | goto out; | |
3186 | } | |
3187 | ||
3188 | if (cc->core_id % smp_threads) { | |
3189 | error_setg(&local_err, "invalid core id %d", cc->core_id); | |
3190 | goto out; | |
3191 | } | |
3192 | ||
459264ef DG |
3193 | /* |
3194 | * In general we should have homogeneous threads-per-core, but old | |
3195 | * (pre hotplug support) machine types allow the last core to have | |
3196 | * reduced threads as a compatibility hack for when we allowed | |
3197 | * total vcpus not a multiple of threads-per-core. | |
3198 | */ | |
3199 | if (mc->has_hotpluggable_cpus && (cc->nr_threads != smp_threads)) { | |
df8658de | 3200 | error_setg(&local_err, "invalid nr-threads %d, must be %d", |
8149e299 | 3201 | cc->nr_threads, smp_threads); |
df8658de | 3202 | goto out; |
8149e299 DG |
3203 | } |
3204 | ||
535455fd IM |
3205 | core_slot = spapr_find_cpu_slot(MACHINE(hotplug_dev), cc->core_id, &index); |
3206 | if (!core_slot) { | |
ff9006dd IM |
3207 | error_setg(&local_err, "core id %d out of range", cc->core_id); |
3208 | goto out; | |
3209 | } | |
3210 | ||
535455fd | 3211 | if (core_slot->cpu) { |
ff9006dd IM |
3212 | error_setg(&local_err, "core %d already populated", cc->core_id); |
3213 | goto out; | |
3214 | } | |
3215 | ||
a0ceb640 | 3216 | numa_cpu_pre_plug(core_slot, dev, &local_err); |
0b8497f0 | 3217 | |
ff9006dd IM |
3218 | out: |
3219 | g_free(base_core_type); | |
3220 | error_propagate(errp, local_err); | |
3221 | } | |
3222 | ||
c20d332a BR |
3223 | static void spapr_machine_device_plug(HotplugHandler *hotplug_dev, |
3224 | DeviceState *dev, Error **errp) | |
3225 | { | |
3226 | sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(qdev_get_machine()); | |
3227 | ||
3228 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { | |
b556854b | 3229 | int node; |
c20d332a BR |
3230 | |
3231 | if (!smc->dr_lmb_enabled) { | |
3232 | error_setg(errp, "Memory hotplug not supported for this machine"); | |
3233 | return; | |
3234 | } | |
9ed442b8 | 3235 | node = object_property_get_uint(OBJECT(dev), PC_DIMM_NODE_PROP, errp); |
c20d332a BR |
3236 | if (*errp) { |
3237 | return; | |
3238 | } | |
1a5512bb GA |
3239 | if (node < 0 || node >= MAX_NODES) { |
3240 | error_setg(errp, "Invaild node %d", node); | |
3241 | return; | |
3242 | } | |
c20d332a | 3243 | |
b556854b BR |
3244 | /* |
3245 | * Currently PowerPC kernel doesn't allow hot-adding memory to | |
3246 | * memory-less node, but instead will silently add the memory | |
3247 | * to the first node that has some memory. This causes two | |
3248 | * unexpected behaviours for the user. | |
3249 | * | |
3250 | * - Memory gets hotplugged to a different node than what the user | |
3251 | * specified. | |
3252 | * - Since pc-dimm subsystem in QEMU still thinks that memory belongs | |
3253 | * to memory-less node, a reboot will set things accordingly | |
3254 | * and the previously hotplugged memory now ends in the right node. | |
3255 | * This appears as if some memory moved from one node to another. | |
3256 | * | |
3257 | * So until kernel starts supporting memory hotplug to memory-less | |
3258 | * nodes, just prevent such attempts upfront in QEMU. | |
3259 | */ | |
3260 | if (nb_numa_nodes && !numa_info[node].node_mem) { | |
3261 | error_setg(errp, "Can't hotplug memory to memory-less node %d", | |
3262 | node); | |
3263 | return; | |
3264 | } | |
3265 | ||
c20d332a | 3266 | spapr_memory_plug(hotplug_dev, dev, node, errp); |
af81cf32 BR |
3267 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { |
3268 | spapr_core_plug(hotplug_dev, dev, errp); | |
c20d332a BR |
3269 | } |
3270 | } | |
3271 | ||
cf632463 BR |
3272 | static void spapr_machine_device_unplug_request(HotplugHandler *hotplug_dev, |
3273 | DeviceState *dev, Error **errp) | |
3274 | { | |
3275 | sPAPRMachineState *sms = SPAPR_MACHINE(qdev_get_machine()); | |
3276 | MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine()); | |
3277 | ||
3278 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { | |
3279 | if (spapr_ovec_test(sms->ov5_cas, OV5_HP_EVT)) { | |
3280 | spapr_memory_unplug_request(hotplug_dev, dev, errp); | |
3281 | } else { | |
3282 | /* NOTE: this means there is a window after guest reset, prior to | |
3283 | * CAS negotiation, where unplug requests will fail due to the | |
3284 | * capability not being detected yet. This is a bit different than | |
3285 | * the case with PCI unplug, where the events will be queued and | |
3286 | * eventually handled by the guest after boot | |
3287 | */ | |
3288 | error_setg(errp, "Memory hot unplug not supported for this guest"); | |
3289 | } | |
6f4b5c3e | 3290 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { |
c5514d0e | 3291 | if (!mc->has_hotpluggable_cpus) { |
6f4b5c3e BR |
3292 | error_setg(errp, "CPU hot unplug not supported on this machine"); |
3293 | return; | |
3294 | } | |
115debf2 | 3295 | spapr_core_unplug_request(hotplug_dev, dev, errp); |
c20d332a BR |
3296 | } |
3297 | } | |
3298 | ||
94a94e4c BR |
3299 | static void spapr_machine_device_pre_plug(HotplugHandler *hotplug_dev, |
3300 | DeviceState *dev, Error **errp) | |
3301 | { | |
c871bc70 LV |
3302 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { |
3303 | spapr_memory_pre_plug(hotplug_dev, dev, errp); | |
3304 | } else if (object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { | |
94a94e4c BR |
3305 | spapr_core_pre_plug(hotplug_dev, dev, errp); |
3306 | } | |
3307 | } | |
3308 | ||
7ebaf795 BR |
3309 | static HotplugHandler *spapr_get_hotplug_handler(MachineState *machine, |
3310 | DeviceState *dev) | |
c20d332a | 3311 | { |
94a94e4c BR |
3312 | if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) || |
3313 | object_dynamic_cast(OBJECT(dev), TYPE_SPAPR_CPU_CORE)) { | |
c20d332a BR |
3314 | return HOTPLUG_HANDLER(machine); |
3315 | } | |
3316 | return NULL; | |
3317 | } | |
3318 | ||
ea089eeb IM |
3319 | static CpuInstanceProperties |
3320 | spapr_cpu_index_to_props(MachineState *machine, unsigned cpu_index) | |
20bb648d | 3321 | { |
ea089eeb IM |
3322 | CPUArchId *core_slot; |
3323 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
3324 | ||
3325 | /* make sure possible_cpu are intialized */ | |
3326 | mc->possible_cpu_arch_ids(machine); | |
3327 | /* get CPU core slot containing thread that matches cpu_index */ | |
3328 | core_slot = spapr_find_cpu_slot(machine, cpu_index, NULL); | |
3329 | assert(core_slot); | |
3330 | return core_slot->props; | |
20bb648d DG |
3331 | } |
3332 | ||
535455fd IM |
3333 | static const CPUArchIdList *spapr_possible_cpu_arch_ids(MachineState *machine) |
3334 | { | |
3335 | int i; | |
3336 | int spapr_max_cores = max_cpus / smp_threads; | |
3337 | MachineClass *mc = MACHINE_GET_CLASS(machine); | |
3338 | ||
c5514d0e | 3339 | if (!mc->has_hotpluggable_cpus) { |
535455fd IM |
3340 | spapr_max_cores = QEMU_ALIGN_UP(smp_cpus, smp_threads) / smp_threads; |
3341 | } | |
3342 | if (machine->possible_cpus) { | |
3343 | assert(machine->possible_cpus->len == spapr_max_cores); | |
3344 | return machine->possible_cpus; | |
3345 | } | |
3346 | ||
3347 | machine->possible_cpus = g_malloc0(sizeof(CPUArchIdList) + | |
3348 | sizeof(CPUArchId) * spapr_max_cores); | |
3349 | machine->possible_cpus->len = spapr_max_cores; | |
3350 | for (i = 0; i < machine->possible_cpus->len; i++) { | |
3351 | int core_id = i * smp_threads; | |
3352 | ||
f2d672c2 | 3353 | machine->possible_cpus->cpus[i].vcpus_count = smp_threads; |
535455fd IM |
3354 | machine->possible_cpus->cpus[i].arch_id = core_id; |
3355 | machine->possible_cpus->cpus[i].props.has_core_id = true; | |
3356 | machine->possible_cpus->cpus[i].props.core_id = core_id; | |
ea089eeb IM |
3357 | |
3358 | /* default distribution of CPUs over NUMA nodes */ | |
3359 | if (nb_numa_nodes) { | |
3360 | /* preset values but do not enable them i.e. 'has_node_id = false', | |
3361 | * numa init code will enable them later if manual mapping wasn't | |
3362 | * present on CLI */ | |
3363 | machine->possible_cpus->cpus[i].props.node_id = | |
3364 | core_id / smp_threads / smp_cores % nb_numa_nodes; | |
3365 | } | |
535455fd IM |
3366 | } |
3367 | return machine->possible_cpus; | |
3368 | } | |
3369 | ||
6737d9ad | 3370 | static void spapr_phb_placement(sPAPRMachineState *spapr, uint32_t index, |
daa23699 DG |
3371 | uint64_t *buid, hwaddr *pio, |
3372 | hwaddr *mmio32, hwaddr *mmio64, | |
6737d9ad DG |
3373 | unsigned n_dma, uint32_t *liobns, Error **errp) |
3374 | { | |
357d1e3b DG |
3375 | /* |
3376 | * New-style PHB window placement. | |
3377 | * | |
3378 | * Goals: Gives large (1TiB), naturally aligned 64-bit MMIO window | |
3379 | * for each PHB, in addition to 2GiB 32-bit MMIO and 64kiB PIO | |
3380 | * windows. | |
3381 | * | |
3382 | * Some guest kernels can't work with MMIO windows above 1<<46 | |
3383 | * (64TiB), so we place up to 31 PHBs in the area 32TiB..64TiB | |
3384 | * | |
3385 | * 32TiB..(33TiB+1984kiB) contains the 64kiB PIO windows for each | |
3386 | * PHB stacked together. (32TiB+2GiB)..(32TiB+64GiB) contains the | |
3387 | * 2GiB 32-bit MMIO windows for each PHB. Then 33..64TiB has the | |
3388 | * 1TiB 64-bit MMIO windows for each PHB. | |
3389 | */ | |
6737d9ad | 3390 | const uint64_t base_buid = 0x800000020000000ULL; |
25e6a118 MT |
3391 | #define SPAPR_MAX_PHBS ((SPAPR_PCI_LIMIT - SPAPR_PCI_BASE) / \ |
3392 | SPAPR_PCI_MEM64_WIN_SIZE - 1) | |
6737d9ad DG |
3393 | int i; |
3394 | ||
357d1e3b DG |
3395 | /* Sanity check natural alignments */ |
3396 | QEMU_BUILD_BUG_ON((SPAPR_PCI_BASE % SPAPR_PCI_MEM64_WIN_SIZE) != 0); | |
3397 | QEMU_BUILD_BUG_ON((SPAPR_PCI_LIMIT % SPAPR_PCI_MEM64_WIN_SIZE) != 0); | |
3398 | QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM64_WIN_SIZE % SPAPR_PCI_MEM32_WIN_SIZE) != 0); | |
3399 | QEMU_BUILD_BUG_ON((SPAPR_PCI_MEM32_WIN_SIZE % SPAPR_PCI_IO_WIN_SIZE) != 0); | |
3400 | /* Sanity check bounds */ | |
25e6a118 MT |
3401 | QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_IO_WIN_SIZE) > |
3402 | SPAPR_PCI_MEM32_WIN_SIZE); | |
3403 | QEMU_BUILD_BUG_ON((SPAPR_MAX_PHBS * SPAPR_PCI_MEM32_WIN_SIZE) > | |
3404 | SPAPR_PCI_MEM64_WIN_SIZE); | |
3405 | ||
3406 | if (index >= SPAPR_MAX_PHBS) { | |
3407 | error_setg(errp, "\"index\" for PAPR PHB is too large (max %llu)", | |
3408 | SPAPR_MAX_PHBS - 1); | |
6737d9ad DG |
3409 | return; |
3410 | } | |
3411 | ||
3412 | *buid = base_buid + index; | |
3413 | for (i = 0; i < n_dma; ++i) { | |
3414 | liobns[i] = SPAPR_PCI_LIOBN(index, i); | |
3415 | } | |
3416 | ||
357d1e3b DG |
3417 | *pio = SPAPR_PCI_BASE + index * SPAPR_PCI_IO_WIN_SIZE; |
3418 | *mmio32 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM32_WIN_SIZE; | |
3419 | *mmio64 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM64_WIN_SIZE; | |
6737d9ad DG |
3420 | } |
3421 | ||
7844e12b CLG |
3422 | static ICSState *spapr_ics_get(XICSFabric *dev, int irq) |
3423 | { | |
3424 | sPAPRMachineState *spapr = SPAPR_MACHINE(dev); | |
3425 | ||
3426 | return ics_valid_irq(spapr->ics, irq) ? spapr->ics : NULL; | |
3427 | } | |
3428 | ||
3429 | static void spapr_ics_resend(XICSFabric *dev) | |
3430 | { | |
3431 | sPAPRMachineState *spapr = SPAPR_MACHINE(dev); | |
3432 | ||
3433 | ics_resend(spapr->ics); | |
3434 | } | |
3435 | ||
81210c20 | 3436 | static ICPState *spapr_icp_get(XICSFabric *xi, int vcpu_id) |
b2fc59aa | 3437 | { |
81210c20 | 3438 | PowerPCCPU *cpu = ppc_get_cpu_by_vcpu_id(vcpu_id); |
b2fc59aa | 3439 | |
5bc8d26d | 3440 | return cpu ? ICP(cpu->intc) : NULL; |
b2fc59aa CLG |
3441 | } |
3442 | ||
6449da45 CLG |
3443 | static void spapr_pic_print_info(InterruptStatsProvider *obj, |
3444 | Monitor *mon) | |
3445 | { | |
3446 | sPAPRMachineState *spapr = SPAPR_MACHINE(obj); | |
5bc8d26d CLG |
3447 | CPUState *cs; |
3448 | ||
3449 | CPU_FOREACH(cs) { | |
3450 | PowerPCCPU *cpu = POWERPC_CPU(cs); | |
6449da45 | 3451 | |
5bc8d26d | 3452 | icp_pic_print_info(ICP(cpu->intc), mon); |
6449da45 CLG |
3453 | } |
3454 | ||
3455 | ics_pic_print_info(spapr->ics, mon); | |
3456 | } | |
3457 | ||
29ee3247 AK |
3458 | static void spapr_machine_class_init(ObjectClass *oc, void *data) |
3459 | { | |
3460 | MachineClass *mc = MACHINE_CLASS(oc); | |
224245bf | 3461 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(oc); |
71461b0f | 3462 | FWPathProviderClass *fwc = FW_PATH_PROVIDER_CLASS(oc); |
34316482 | 3463 | NMIClass *nc = NMI_CLASS(oc); |
c20d332a | 3464 | HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc); |
1d1be34d | 3465 | PPCVirtualHypervisorClass *vhc = PPC_VIRTUAL_HYPERVISOR_CLASS(oc); |
7844e12b | 3466 | XICSFabricClass *xic = XICS_FABRIC_CLASS(oc); |
6449da45 | 3467 | InterruptStatsProviderClass *ispc = INTERRUPT_STATS_PROVIDER_CLASS(oc); |
958db90c | 3468 | |
0eb9054c | 3469 | mc->desc = "pSeries Logical Partition (PAPR compliant)"; |
fc9f38c3 DG |
3470 | |
3471 | /* | |
3472 | * We set up the default / latest behaviour here. The class_init | |
3473 | * functions for the specific versioned machine types can override | |
3474 | * these details for backwards compatibility | |
3475 | */ | |
958db90c MA |
3476 | mc->init = ppc_spapr_init; |
3477 | mc->reset = ppc_spapr_reset; | |
3478 | mc->block_default_type = IF_SCSI; | |
6244bb7e | 3479 | mc->max_cpus = 1024; |
958db90c | 3480 | mc->no_parallel = 1; |
5b2128d2 | 3481 | mc->default_boot_order = ""; |
a34944fe | 3482 | mc->default_ram_size = 512 * M_BYTE; |
958db90c | 3483 | mc->kvm_type = spapr_kvm_type; |
9e3f9733 | 3484 | mc->has_dynamic_sysbus = true; |
e4024630 | 3485 | mc->pci_allow_0_address = true; |
7ebaf795 | 3486 | mc->get_hotplug_handler = spapr_get_hotplug_handler; |
94a94e4c | 3487 | hc->pre_plug = spapr_machine_device_pre_plug; |
c20d332a | 3488 | hc->plug = spapr_machine_device_plug; |
ea089eeb | 3489 | mc->cpu_index_to_instance_props = spapr_cpu_index_to_props; |
535455fd | 3490 | mc->possible_cpu_arch_ids = spapr_possible_cpu_arch_ids; |
cf632463 | 3491 | hc->unplug_request = spapr_machine_device_unplug_request; |
00b4fbe2 | 3492 | |
fc9f38c3 | 3493 | smc->dr_lmb_enabled = true; |
3daa4a9f | 3494 | smc->tcg_default_cpu = "POWER8"; |
c5514d0e | 3495 | mc->has_hotpluggable_cpus = true; |
52b81ab5 | 3496 | smc->resize_hpt_default = SPAPR_RESIZE_HPT_ENABLED; |
71461b0f | 3497 | fwc->get_dev_path = spapr_get_fw_dev_path; |
34316482 | 3498 | nc->nmi_monitor_handler = spapr_nmi; |
6737d9ad | 3499 | smc->phb_placement = spapr_phb_placement; |
1d1be34d | 3500 | vhc->hypercall = emulate_spapr_hypercall; |
e57ca75c DG |
3501 | vhc->hpt_mask = spapr_hpt_mask; |
3502 | vhc->map_hptes = spapr_map_hptes; | |
3503 | vhc->unmap_hptes = spapr_unmap_hptes; | |
3504 | vhc->store_hpte = spapr_store_hpte; | |
9861bb3e | 3505 | vhc->get_patbe = spapr_get_patbe; |
7844e12b CLG |
3506 | xic->ics_get = spapr_ics_get; |
3507 | xic->ics_resend = spapr_ics_resend; | |
b2fc59aa | 3508 | xic->icp_get = spapr_icp_get; |
6449da45 | 3509 | ispc->print_info = spapr_pic_print_info; |
55641213 LV |
3510 | /* Force NUMA node memory size to be a multiple of |
3511 | * SPAPR_MEMORY_BLOCK_SIZE (256M) since that's the granularity | |
3512 | * in which LMBs are represented and hot-added | |
3513 | */ | |
3514 | mc->numa_mem_align_shift = 28; | |
29ee3247 AK |
3515 | } |
3516 | ||
3517 | static const TypeInfo spapr_machine_info = { | |
3518 | .name = TYPE_SPAPR_MACHINE, | |
3519 | .parent = TYPE_MACHINE, | |
4aee7362 | 3520 | .abstract = true, |
6ca1502e | 3521 | .instance_size = sizeof(sPAPRMachineState), |
23825581 | 3522 | .instance_init = spapr_machine_initfn, |
87bbdd9c | 3523 | .instance_finalize = spapr_machine_finalizefn, |
183930c0 | 3524 | .class_size = sizeof(sPAPRMachineClass), |
29ee3247 | 3525 | .class_init = spapr_machine_class_init, |
71461b0f AK |
3526 | .interfaces = (InterfaceInfo[]) { |
3527 | { TYPE_FW_PATH_PROVIDER }, | |
34316482 | 3528 | { TYPE_NMI }, |
c20d332a | 3529 | { TYPE_HOTPLUG_HANDLER }, |
1d1be34d | 3530 | { TYPE_PPC_VIRTUAL_HYPERVISOR }, |
7844e12b | 3531 | { TYPE_XICS_FABRIC }, |
6449da45 | 3532 | { TYPE_INTERRUPT_STATS_PROVIDER }, |
71461b0f AK |
3533 | { } |
3534 | }, | |
29ee3247 AK |
3535 | }; |
3536 | ||
fccbc785 | 3537 | #define DEFINE_SPAPR_MACHINE(suffix, verstr, latest) \ |
5013c547 DG |
3538 | static void spapr_machine_##suffix##_class_init(ObjectClass *oc, \ |
3539 | void *data) \ | |
3540 | { \ | |
3541 | MachineClass *mc = MACHINE_CLASS(oc); \ | |
3542 | spapr_machine_##suffix##_class_options(mc); \ | |
fccbc785 DG |
3543 | if (latest) { \ |
3544 | mc->alias = "pseries"; \ | |
3545 | mc->is_default = 1; \ | |
3546 | } \ | |
5013c547 DG |
3547 | } \ |
3548 | static void spapr_machine_##suffix##_instance_init(Object *obj) \ | |
3549 | { \ | |
3550 | MachineState *machine = MACHINE(obj); \ | |
3551 | spapr_machine_##suffix##_instance_options(machine); \ | |
3552 | } \ | |
3553 | static const TypeInfo spapr_machine_##suffix##_info = { \ | |
3554 | .name = MACHINE_TYPE_NAME("pseries-" verstr), \ | |
3555 | .parent = TYPE_SPAPR_MACHINE, \ | |
3556 | .class_init = spapr_machine_##suffix##_class_init, \ | |
3557 | .instance_init = spapr_machine_##suffix##_instance_init, \ | |
3558 | }; \ | |
3559 | static void spapr_machine_register_##suffix(void) \ | |
3560 | { \ | |
3561 | type_register(&spapr_machine_##suffix##_info); \ | |
3562 | } \ | |
0e6aac87 | 3563 | type_init(spapr_machine_register_##suffix) |
5013c547 | 3564 | |
e2676b16 GK |
3565 | /* |
3566 | * pseries-2.11 | |
3567 | */ | |
3568 | static void spapr_machine_2_11_instance_options(MachineState *machine) | |
3569 | { | |
3570 | } | |
3571 | ||
3572 | static void spapr_machine_2_11_class_options(MachineClass *mc) | |
3573 | { | |
3574 | /* Defaults for the latest behaviour inherited from the base class */ | |
3575 | } | |
3576 | ||
3577 | DEFINE_SPAPR_MACHINE(2_11, "2.11", true); | |
3578 | ||
3fa14fbe DG |
3579 | /* |
3580 | * pseries-2.10 | |
3581 | */ | |
e2676b16 GK |
3582 | #define SPAPR_COMPAT_2_10 \ |
3583 | HW_COMPAT_2_10 \ | |
3584 | ||
3fa14fbe DG |
3585 | static void spapr_machine_2_10_instance_options(MachineState *machine) |
3586 | { | |
3587 | } | |
3588 | ||
3589 | static void spapr_machine_2_10_class_options(MachineClass *mc) | |
3590 | { | |
e2676b16 GK |
3591 | spapr_machine_2_11_class_options(mc); |
3592 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_10); | |
3fa14fbe DG |
3593 | } |
3594 | ||
e2676b16 | 3595 | DEFINE_SPAPR_MACHINE(2_10, "2.10", false); |
3fa14fbe | 3596 | |
fa325e6c DG |
3597 | /* |
3598 | * pseries-2.9 | |
3599 | */ | |
3fa14fbe | 3600 | #define SPAPR_COMPAT_2_9 \ |
d5fc133e DG |
3601 | HW_COMPAT_2_9 \ |
3602 | { \ | |
3603 | .driver = TYPE_POWERPC_CPU, \ | |
3604 | .property = "pre-2.10-migration", \ | |
3605 | .value = "on", \ | |
3606 | }, \ | |
3fa14fbe | 3607 | |
fa325e6c DG |
3608 | static void spapr_machine_2_9_instance_options(MachineState *machine) |
3609 | { | |
3fa14fbe | 3610 | spapr_machine_2_10_instance_options(machine); |
fa325e6c DG |
3611 | } |
3612 | ||
3613 | static void spapr_machine_2_9_class_options(MachineClass *mc) | |
3614 | { | |
46f7afa3 GK |
3615 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3616 | ||
3fa14fbe DG |
3617 | spapr_machine_2_10_class_options(mc); |
3618 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_9); | |
3bfe5716 | 3619 | mc->numa_auto_assign_ram = numa_legacy_auto_assign_ram; |
46f7afa3 | 3620 | smc->pre_2_10_has_unused_icps = true; |
52b81ab5 | 3621 | smc->resize_hpt_default = SPAPR_RESIZE_HPT_DISABLED; |
fa325e6c DG |
3622 | } |
3623 | ||
3fa14fbe | 3624 | DEFINE_SPAPR_MACHINE(2_9, "2.9", false); |
fa325e6c | 3625 | |
db800b21 DG |
3626 | /* |
3627 | * pseries-2.8 | |
3628 | */ | |
82516263 DG |
3629 | #define SPAPR_COMPAT_2_8 \ |
3630 | HW_COMPAT_2_8 \ | |
3631 | { \ | |
3632 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3633 | .property = "pcie-extended-configuration-space", \ | |
3634 | .value = "off", \ | |
3635 | }, | |
fa325e6c | 3636 | |
db800b21 DG |
3637 | static void spapr_machine_2_8_instance_options(MachineState *machine) |
3638 | { | |
fa325e6c | 3639 | spapr_machine_2_9_instance_options(machine); |
db800b21 DG |
3640 | } |
3641 | ||
3642 | static void spapr_machine_2_8_class_options(MachineClass *mc) | |
3643 | { | |
fa325e6c DG |
3644 | spapr_machine_2_9_class_options(mc); |
3645 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_8); | |
55641213 | 3646 | mc->numa_mem_align_shift = 23; |
db800b21 DG |
3647 | } |
3648 | ||
fa325e6c | 3649 | DEFINE_SPAPR_MACHINE(2_8, "2.8", false); |
db800b21 | 3650 | |
1ea1eefc BR |
3651 | /* |
3652 | * pseries-2.7 | |
3653 | */ | |
357d1e3b DG |
3654 | #define SPAPR_COMPAT_2_7 \ |
3655 | HW_COMPAT_2_7 \ | |
3656 | { \ | |
3657 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3658 | .property = "mem_win_size", \ | |
3659 | .value = stringify(SPAPR_PCI_2_7_MMIO_WIN_SIZE),\ | |
3660 | }, \ | |
3661 | { \ | |
3662 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3663 | .property = "mem64_win_size", \ | |
3664 | .value = "0", \ | |
146c11f1 DG |
3665 | }, \ |
3666 | { \ | |
3667 | .driver = TYPE_POWERPC_CPU, \ | |
3668 | .property = "pre-2.8-migration", \ | |
3669 | .value = "on", \ | |
5c4537bd DG |
3670 | }, \ |
3671 | { \ | |
3672 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE, \ | |
3673 | .property = "pre-2.8-migration", \ | |
3674 | .value = "on", \ | |
357d1e3b DG |
3675 | }, |
3676 | ||
3677 | static void phb_placement_2_7(sPAPRMachineState *spapr, uint32_t index, | |
3678 | uint64_t *buid, hwaddr *pio, | |
3679 | hwaddr *mmio32, hwaddr *mmio64, | |
3680 | unsigned n_dma, uint32_t *liobns, Error **errp) | |
3681 | { | |
3682 | /* Legacy PHB placement for pseries-2.7 and earlier machine types */ | |
3683 | const uint64_t base_buid = 0x800000020000000ULL; | |
3684 | const hwaddr phb_spacing = 0x1000000000ULL; /* 64 GiB */ | |
3685 | const hwaddr mmio_offset = 0xa0000000; /* 2 GiB + 512 MiB */ | |
3686 | const hwaddr pio_offset = 0x80000000; /* 2 GiB */ | |
3687 | const uint32_t max_index = 255; | |
3688 | const hwaddr phb0_alignment = 0x10000000000ULL; /* 1 TiB */ | |
3689 | ||
3690 | uint64_t ram_top = MACHINE(spapr)->ram_size; | |
3691 | hwaddr phb0_base, phb_base; | |
3692 | int i; | |
3693 | ||
3694 | /* Do we have hotpluggable memory? */ | |
3695 | if (MACHINE(spapr)->maxram_size > ram_top) { | |
3696 | /* Can't just use maxram_size, because there may be an | |
3697 | * alignment gap between normal and hotpluggable memory | |
3698 | * regions */ | |
3699 | ram_top = spapr->hotplug_memory.base + | |
3700 | memory_region_size(&spapr->hotplug_memory.mr); | |
3701 | } | |
3702 | ||
3703 | phb0_base = QEMU_ALIGN_UP(ram_top, phb0_alignment); | |
3704 | ||
3705 | if (index > max_index) { | |
3706 | error_setg(errp, "\"index\" for PAPR PHB is too large (max %u)", | |
3707 | max_index); | |
3708 | return; | |
3709 | } | |
3710 | ||
3711 | *buid = base_buid + index; | |
3712 | for (i = 0; i < n_dma; ++i) { | |
3713 | liobns[i] = SPAPR_PCI_LIOBN(index, i); | |
3714 | } | |
3715 | ||
3716 | phb_base = phb0_base + index * phb_spacing; | |
3717 | *pio = phb_base + pio_offset; | |
3718 | *mmio32 = phb_base + mmio_offset; | |
3719 | /* | |
3720 | * We don't set the 64-bit MMIO window, relying on the PHB's | |
3721 | * fallback behaviour of automatically splitting a large "32-bit" | |
3722 | * window into contiguous 32-bit and 64-bit windows | |
3723 | */ | |
3724 | } | |
db800b21 | 3725 | |
1ea1eefc BR |
3726 | static void spapr_machine_2_7_instance_options(MachineState *machine) |
3727 | { | |
f6229214 MR |
3728 | sPAPRMachineState *spapr = SPAPR_MACHINE(machine); |
3729 | ||
672de881 | 3730 | spapr_machine_2_8_instance_options(machine); |
f6229214 | 3731 | spapr->use_hotplug_event_source = false; |
1ea1eefc BR |
3732 | } |
3733 | ||
3734 | static void spapr_machine_2_7_class_options(MachineClass *mc) | |
3735 | { | |
3daa4a9f TH |
3736 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3737 | ||
db800b21 | 3738 | spapr_machine_2_8_class_options(mc); |
3daa4a9f | 3739 | smc->tcg_default_cpu = "POWER7"; |
db800b21 | 3740 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_7); |
357d1e3b | 3741 | smc->phb_placement = phb_placement_2_7; |
1ea1eefc BR |
3742 | } |
3743 | ||
db800b21 | 3744 | DEFINE_SPAPR_MACHINE(2_7, "2.7", false); |
1ea1eefc | 3745 | |
4b23699c DG |
3746 | /* |
3747 | * pseries-2.6 | |
3748 | */ | |
1ea1eefc | 3749 | #define SPAPR_COMPAT_2_6 \ |
ae4de14c AK |
3750 | HW_COMPAT_2_6 \ |
3751 | { \ | |
3752 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ | |
3753 | .property = "ddw",\ | |
3754 | .value = stringify(off),\ | |
3755 | }, | |
1ea1eefc | 3756 | |
4b23699c DG |
3757 | static void spapr_machine_2_6_instance_options(MachineState *machine) |
3758 | { | |
672de881 | 3759 | spapr_machine_2_7_instance_options(machine); |
4b23699c DG |
3760 | } |
3761 | ||
3762 | static void spapr_machine_2_6_class_options(MachineClass *mc) | |
3763 | { | |
1ea1eefc | 3764 | spapr_machine_2_7_class_options(mc); |
c5514d0e | 3765 | mc->has_hotpluggable_cpus = false; |
1ea1eefc | 3766 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_6); |
4b23699c DG |
3767 | } |
3768 | ||
1ea1eefc | 3769 | DEFINE_SPAPR_MACHINE(2_6, "2.6", false); |
4b23699c | 3770 | |
1c5f29bb DG |
3771 | /* |
3772 | * pseries-2.5 | |
3773 | */ | |
4b23699c | 3774 | #define SPAPR_COMPAT_2_5 \ |
57c522f4 TH |
3775 | HW_COMPAT_2_5 \ |
3776 | { \ | |
3777 | .driver = "spapr-vlan", \ | |
3778 | .property = "use-rx-buffer-pools", \ | |
3779 | .value = "off", \ | |
3780 | }, | |
4b23699c | 3781 | |
5013c547 | 3782 | static void spapr_machine_2_5_instance_options(MachineState *machine) |
1c5f29bb | 3783 | { |
672de881 | 3784 | spapr_machine_2_6_instance_options(machine); |
5013c547 DG |
3785 | } |
3786 | ||
3787 | static void spapr_machine_2_5_class_options(MachineClass *mc) | |
3788 | { | |
57040d45 TH |
3789 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3790 | ||
4b23699c | 3791 | spapr_machine_2_6_class_options(mc); |
57040d45 | 3792 | smc->use_ohci_by_default = true; |
4b23699c | 3793 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_5); |
1c5f29bb DG |
3794 | } |
3795 | ||
4b23699c | 3796 | DEFINE_SPAPR_MACHINE(2_5, "2.5", false); |
1c5f29bb DG |
3797 | |
3798 | /* | |
3799 | * pseries-2.4 | |
3800 | */ | |
80fd50f9 CH |
3801 | #define SPAPR_COMPAT_2_4 \ |
3802 | HW_COMPAT_2_4 | |
3803 | ||
5013c547 | 3804 | static void spapr_machine_2_4_instance_options(MachineState *machine) |
1c5f29bb | 3805 | { |
5013c547 DG |
3806 | spapr_machine_2_5_instance_options(machine); |
3807 | } | |
1c5f29bb | 3808 | |
5013c547 DG |
3809 | static void spapr_machine_2_4_class_options(MachineClass *mc) |
3810 | { | |
fc9f38c3 DG |
3811 | sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc); |
3812 | ||
3813 | spapr_machine_2_5_class_options(mc); | |
fc9f38c3 | 3814 | smc->dr_lmb_enabled = false; |
f949b4e5 | 3815 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_4); |
1c5f29bb DG |
3816 | } |
3817 | ||
fccbc785 | 3818 | DEFINE_SPAPR_MACHINE(2_4, "2.4", false); |
1c5f29bb DG |
3819 | |
3820 | /* | |
3821 | * pseries-2.3 | |
3822 | */ | |
38ff32c6 | 3823 | #define SPAPR_COMPAT_2_3 \ |
7619c7b0 MR |
3824 | HW_COMPAT_2_3 \ |
3825 | {\ | |
3826 | .driver = "spapr-pci-host-bridge",\ | |
3827 | .property = "dynamic-reconfiguration",\ | |
3828 | .value = "off",\ | |
3829 | }, | |
38ff32c6 | 3830 | |
5013c547 | 3831 | static void spapr_machine_2_3_instance_options(MachineState *machine) |
d25228e7 | 3832 | { |
5013c547 | 3833 | spapr_machine_2_4_instance_options(machine); |
d25228e7 JW |
3834 | } |
3835 | ||
5013c547 | 3836 | static void spapr_machine_2_3_class_options(MachineClass *mc) |
6026db45 | 3837 | { |
fc9f38c3 | 3838 | spapr_machine_2_4_class_options(mc); |
f949b4e5 | 3839 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_3); |
6026db45 | 3840 | } |
fccbc785 | 3841 | DEFINE_SPAPR_MACHINE(2_3, "2.3", false); |
6026db45 | 3842 | |
1c5f29bb DG |
3843 | /* |
3844 | * pseries-2.2 | |
3845 | */ | |
3846 | ||
3847 | #define SPAPR_COMPAT_2_2 \ | |
1c5f29bb DG |
3848 | HW_COMPAT_2_2 \ |
3849 | {\ | |
3850 | .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\ | |
3851 | .property = "mem_win_size",\ | |
3852 | .value = "0x20000000",\ | |
3853 | }, | |
3854 | ||
5013c547 | 3855 | static void spapr_machine_2_2_instance_options(MachineState *machine) |
1c5f29bb | 3856 | { |
5013c547 | 3857 | spapr_machine_2_3_instance_options(machine); |
cba0e779 | 3858 | machine->suppress_vmdesc = true; |
1c5f29bb DG |
3859 | } |
3860 | ||
5013c547 | 3861 | static void spapr_machine_2_2_class_options(MachineClass *mc) |
4aee7362 | 3862 | { |
fc9f38c3 | 3863 | spapr_machine_2_3_class_options(mc); |
f949b4e5 | 3864 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_2); |
4aee7362 | 3865 | } |
fccbc785 | 3866 | DEFINE_SPAPR_MACHINE(2_2, "2.2", false); |
4aee7362 | 3867 | |
1c5f29bb DG |
3868 | /* |
3869 | * pseries-2.1 | |
3870 | */ | |
3871 | #define SPAPR_COMPAT_2_1 \ | |
1c5f29bb | 3872 | HW_COMPAT_2_1 |
3dab0244 | 3873 | |
5013c547 | 3874 | static void spapr_machine_2_1_instance_options(MachineState *machine) |
1c5f29bb | 3875 | { |
5013c547 | 3876 | spapr_machine_2_2_instance_options(machine); |
1c5f29bb | 3877 | } |
d25228e7 | 3878 | |
5013c547 | 3879 | static void spapr_machine_2_1_class_options(MachineClass *mc) |
d25228e7 | 3880 | { |
fc9f38c3 | 3881 | spapr_machine_2_2_class_options(mc); |
f949b4e5 | 3882 | SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_1); |
d25228e7 | 3883 | } |
fccbc785 | 3884 | DEFINE_SPAPR_MACHINE(2_1, "2.1", false); |
fb0fc8f6 | 3885 | |
29ee3247 | 3886 | static void spapr_machine_register_types(void) |
9fdf0c29 | 3887 | { |
29ee3247 | 3888 | type_register_static(&spapr_machine_info); |
9fdf0c29 DG |
3889 | } |
3890 | ||
29ee3247 | 3891 | type_init(spapr_machine_register_types) |