2 * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
4 * Hypercall based emulated RTAS
6 * Copyright (c) 2010-2011 David Gibson, IBM Corporation.
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:
15 * The above copyright notice and this permission notice shall be included in
16 * all copies or substantial portions of the Software.
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
28 #include "qemu/osdep.h"
31 #include "qemu/error-report.h"
32 #include "sysemu/sysemu.h"
33 #include "sysemu/device_tree.h"
34 #include "sysemu/cpus.h"
35 #include "sysemu/hw_accel.h"
36 #include "sysemu/runstate.h"
39 #include "hw/ppc/spapr.h"
40 #include "hw/ppc/spapr_vio.h"
41 #include "hw/ppc/spapr_rtas.h"
42 #include "hw/ppc/spapr_cpu_core.h"
43 #include "hw/ppc/ppc.h"
44 #include "hw/boards.h"
47 #include "hw/ppc/spapr_drc.h"
48 #include "qemu/cutils.h"
50 #include "hw/ppc/fdt.h"
51 #include "target/ppc/mmu-hash64.h"
52 #include "target/ppc/mmu-book3s-v3.h"
53 #include "migration/blocker.h"
55 static void rtas_display_character(PowerPCCPU *cpu, SpaprMachineState *spapr,
56 uint32_t token, uint32_t nargs,
58 uint32_t nret, target_ulong rets)
60 uint8_t c = rtas_ld(args, 0);
61 SpaprVioDevice *sdev = vty_lookup(spapr, 0);
64 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
66 vty_putchars(sdev, &c, sizeof(c));
67 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
71 static void rtas_power_off(PowerPCCPU *cpu, SpaprMachineState *spapr,
72 uint32_t token, uint32_t nargs, target_ulong args,
73 uint32_t nret, target_ulong rets)
75 if (nargs != 2 || nret != 1) {
76 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
79 qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
81 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
84 static void rtas_system_reboot(PowerPCCPU *cpu, SpaprMachineState *spapr,
85 uint32_t token, uint32_t nargs,
87 uint32_t nret, target_ulong rets)
89 if (nargs != 0 || nret != 1) {
90 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
93 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
94 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
97 static void rtas_query_cpu_stopped_state(PowerPCCPU *cpu_,
98 SpaprMachineState *spapr,
99 uint32_t token, uint32_t nargs,
101 uint32_t nret, target_ulong rets)
106 if (nargs != 1 || nret != 2) {
107 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
111 id = rtas_ld(args, 0);
112 cpu = spapr_find_cpu(id);
114 if (CPU(cpu)->halted) {
120 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
124 /* Didn't find a matching cpu */
125 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
128 static void rtas_start_cpu(PowerPCCPU *callcpu, SpaprMachineState *spapr,
129 uint32_t token, uint32_t nargs,
131 uint32_t nret, target_ulong rets)
133 target_ulong id, start, r3;
136 PowerPCCPUClass *pcc;
139 if (nargs != 3 || nret != 1) {
140 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
144 id = rtas_ld(args, 0);
145 start = rtas_ld(args, 1);
146 r3 = rtas_ld(args, 2);
148 newcpu = spapr_find_cpu(id);
150 /* Didn't find a matching cpu */
151 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
156 pcc = POWERPC_CPU_GET_CLASS(newcpu);
158 if (!CPU(newcpu)->halted) {
159 rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
163 cpu_synchronize_state(CPU(newcpu));
165 env->msr = (1ULL << MSR_SF) | (1ULL << MSR_ME);
167 /* Enable Power-saving mode Exit Cause exceptions for the new CPU */
168 lpcr = env->spr[SPR_LPCR];
169 if (!pcc->interrupts_big_endian(callcpu)) {
172 if (env->mmu_model == POWERPC_MMU_3_00) {
174 * New cpus are expected to start in the same radix/hash mode
175 * as the existing CPUs
177 if (ppc64_v3_radix(callcpu)) {
178 lpcr |= LPCR_UPRT | LPCR_GTSE | LPCR_HR;
180 lpcr &= ~(LPCR_UPRT | LPCR_GTSE | LPCR_HR);
182 env->spr[SPR_PSSCR] &= ~PSSCR_EC;
184 ppc_store_lpcr(newcpu, lpcr);
187 * Set the timebase offset of the new CPU to that of the invoking
188 * CPU. This helps hotplugged CPU to have the correct timebase
191 newcpu->env.tb_env->tb_offset = callcpu->env.tb_env->tb_offset;
193 spapr_cpu_set_entry_state(newcpu, start, 0, r3, 0);
195 qemu_cpu_kick(CPU(newcpu));
197 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
200 static void rtas_stop_self(PowerPCCPU *cpu, SpaprMachineState *spapr,
201 uint32_t token, uint32_t nargs,
203 uint32_t nret, target_ulong rets)
205 CPUState *cs = CPU(cpu);
206 CPUPPCState *env = &cpu->env;
207 PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
209 /* Disable Power-saving mode Exit Cause exceptions for the CPU.
210 * This could deliver an interrupt on a dying CPU and crash the
212 * For the same reason, set PSSCR_EC.
214 ppc_store_lpcr(cpu, env->spr[SPR_LPCR] & ~pcc->lpcr_pm);
215 env->spr[SPR_PSSCR] |= PSSCR_EC;
217 kvmppc_set_reg_ppc_online(cpu, 0);
221 static void rtas_ibm_suspend_me(PowerPCCPU *cpu, SpaprMachineState *spapr,
222 uint32_t token, uint32_t nargs,
224 uint32_t nret, target_ulong rets)
228 if (nargs != 0 || nret != 1) {
229 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
234 PowerPCCPU *c = POWERPC_CPU(cs);
235 CPUPPCState *e = &c->env;
241 if (!cs->halted || (e->msr & (1ULL << MSR_EE))) {
242 rtas_st(rets, 0, H_MULTI_THREADS_ACTIVE);
247 qemu_system_suspend_request();
248 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
251 static inline int sysparm_st(target_ulong addr, target_ulong len,
252 const void *val, uint16_t vallen)
254 hwaddr phys = ppc64_phys_to_real(addr);
257 return RTAS_OUT_SYSPARM_PARAM_ERROR;
259 stw_be_phys(&address_space_memory, phys, vallen);
260 cpu_physical_memory_write(phys + 2, val, MIN(len - 2, vallen));
261 return RTAS_OUT_SUCCESS;
264 static void rtas_ibm_get_system_parameter(PowerPCCPU *cpu,
265 SpaprMachineState *spapr,
266 uint32_t token, uint32_t nargs,
268 uint32_t nret, target_ulong rets)
270 PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cpu);
271 MachineState *ms = MACHINE(spapr);
272 target_ulong parameter = rtas_ld(args, 0);
273 target_ulong buffer = rtas_ld(args, 1);
274 target_ulong length = rtas_ld(args, 2);
278 case RTAS_SYSPARM_SPLPAR_CHARACTERISTICS: {
279 char *param_val = g_strdup_printf("MaxEntCap=%d,"
280 "DesMem=%" PRIu64 ","
287 if (pcc->n_host_threads > 0) {
288 char *hostthr_val, *old = param_val;
291 * Add HostThrs property. This property is not present in PAPR but
292 * is expected by some guests to communicate the number of physical
293 * host threads per core on the system so that they can scale
294 * information which varies based on the thread configuration.
296 hostthr_val = g_strdup_printf(",HostThrs=%d", pcc->n_host_threads);
297 param_val = g_strconcat(param_val, hostthr_val, NULL);
301 ret = sysparm_st(buffer, length, param_val, strlen(param_val) + 1);
305 case RTAS_SYSPARM_DIAGNOSTICS_RUN_MODE: {
306 uint8_t param_val = DIAGNOSTICS_RUN_MODE_DISABLED;
308 ret = sysparm_st(buffer, length, ¶m_val, sizeof(param_val));
311 case RTAS_SYSPARM_UUID:
312 ret = sysparm_st(buffer, length, (unsigned char *)&qemu_uuid,
313 (qemu_uuid_set ? 16 : 0));
316 ret = RTAS_OUT_NOT_SUPPORTED;
319 rtas_st(rets, 0, ret);
322 static void rtas_ibm_set_system_parameter(PowerPCCPU *cpu,
323 SpaprMachineState *spapr,
324 uint32_t token, uint32_t nargs,
326 uint32_t nret, target_ulong rets)
328 target_ulong parameter = rtas_ld(args, 0);
329 target_ulong ret = RTAS_OUT_NOT_SUPPORTED;
332 case RTAS_SYSPARM_SPLPAR_CHARACTERISTICS:
333 case RTAS_SYSPARM_DIAGNOSTICS_RUN_MODE:
334 case RTAS_SYSPARM_UUID:
335 ret = RTAS_OUT_NOT_AUTHORIZED;
339 rtas_st(rets, 0, ret);
342 static void rtas_ibm_os_term(PowerPCCPU *cpu,
343 SpaprMachineState *spapr,
344 uint32_t token, uint32_t nargs,
346 uint32_t nret, target_ulong rets)
348 target_ulong msgaddr = rtas_ld(args, 0);
351 cpu_physical_memory_read(msgaddr, msg, sizeof(msg) - 1);
352 msg[sizeof(msg) - 1] = 0;
354 error_report("OS terminated: %s", msg);
355 qemu_system_guest_panicked(NULL);
357 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
360 static void rtas_set_power_level(PowerPCCPU *cpu, SpaprMachineState *spapr,
361 uint32_t token, uint32_t nargs,
362 target_ulong args, uint32_t nret,
365 int32_t power_domain;
367 if (nargs != 2 || nret != 2) {
368 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
372 /* we currently only use a single, "live insert" powerdomain for
373 * hotplugged/dlpar'd resources, so the power is always live/full (100)
375 power_domain = rtas_ld(args, 0);
376 if (power_domain != -1) {
377 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
381 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
382 rtas_st(rets, 1, 100);
385 static void rtas_get_power_level(PowerPCCPU *cpu, SpaprMachineState *spapr,
386 uint32_t token, uint32_t nargs,
387 target_ulong args, uint32_t nret,
390 int32_t power_domain;
392 if (nargs != 1 || nret != 2) {
393 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
397 /* we currently only use a single, "live insert" powerdomain for
398 * hotplugged/dlpar'd resources, so the power is always live/full (100)
400 power_domain = rtas_ld(args, 0);
401 if (power_domain != -1) {
402 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
406 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
407 rtas_st(rets, 1, 100);
410 static void rtas_ibm_nmi_register(PowerPCCPU *cpu,
411 SpaprMachineState *spapr,
412 uint32_t token, uint32_t nargs,
414 uint32_t nret, target_ulong rets)
417 target_ulong sreset_addr, mce_addr;
419 if (spapr_get_cap(spapr, SPAPR_CAP_FWNMI) == SPAPR_CAP_OFF) {
420 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
424 rtas_addr = spapr_get_rtas_addr();
426 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
430 sreset_addr = rtas_ld(args, 0);
431 mce_addr = rtas_ld(args, 1);
433 /* PAPR requires these are in the first 32M of memory and within RMA */
434 if (sreset_addr >= 32 * MiB || sreset_addr >= spapr->rma_size ||
435 mce_addr >= 32 * MiB || mce_addr >= spapr->rma_size) {
436 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
441 if (kvmppc_set_fwnmi(cpu) < 0) {
442 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
447 spapr->fwnmi_system_reset_addr = sreset_addr;
448 spapr->fwnmi_machine_check_addr = mce_addr;
450 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
453 static void rtas_ibm_nmi_interlock(PowerPCCPU *cpu,
454 SpaprMachineState *spapr,
455 uint32_t token, uint32_t nargs,
457 uint32_t nret, target_ulong rets)
459 if (spapr_get_cap(spapr, SPAPR_CAP_FWNMI) == SPAPR_CAP_OFF) {
460 rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
464 if (spapr->fwnmi_machine_check_addr == -1) {
465 qemu_log_mask(LOG_GUEST_ERROR,
466 "FWNMI: ibm,nmi-interlock RTAS called with FWNMI not registered.\n");
468 /* NMI register not called */
469 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
473 if (spapr->fwnmi_machine_check_interlock != cpu->vcpu_id) {
475 * The vCPU that hit the NMI should invoke "ibm,nmi-interlock"
476 * This should be PARAM_ERROR, but Linux calls "ibm,nmi-interlock"
477 * for system reset interrupts, despite them not being interlocked.
478 * PowerVM silently ignores this and returns success here. Returning
479 * failure causes Linux to print the error "FWNMI: nmi-interlock
480 * failed: -3", although no other apparent ill effects, this is a
481 * regression for the user when enabling FWNMI. So for now, match
482 * PowerVM. When most Linux clients are fixed, this could be
485 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
490 * vCPU issuing "ibm,nmi-interlock" is done with NMI handling,
491 * hence unset fwnmi_machine_check_interlock.
493 spapr->fwnmi_machine_check_interlock = -1;
494 qemu_cond_signal(&spapr->fwnmi_machine_check_interlock_cond);
495 rtas_st(rets, 0, RTAS_OUT_SUCCESS);
496 migrate_del_blocker(spapr->fwnmi_migration_blocker);
499 static struct rtas_call {
502 } rtas_table[RTAS_TOKEN_MAX - RTAS_TOKEN_BASE];
504 target_ulong spapr_rtas_call(PowerPCCPU *cpu, SpaprMachineState *spapr,
505 uint32_t token, uint32_t nargs, target_ulong args,
506 uint32_t nret, target_ulong rets)
508 if ((token >= RTAS_TOKEN_BASE) && (token < RTAS_TOKEN_MAX)) {
509 struct rtas_call *call = rtas_table + (token - RTAS_TOKEN_BASE);
512 call->fn(cpu, spapr, token, nargs, args, nret, rets);
517 /* HACK: Some Linux early debug code uses RTAS display-character,
518 * but assumes the token value is 0xa (which it is on some real
519 * machines) without looking it up in the device tree. This
520 * special case makes this work */
522 rtas_display_character(cpu, spapr, 0xa, nargs, args, nret, rets);
526 hcall_dprintf("Unknown RTAS token 0x%x\n", token);
527 rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
531 uint64_t qtest_rtas_call(char *cmd, uint32_t nargs, uint64_t args,
532 uint32_t nret, uint64_t rets)
536 for (token = 0; token < RTAS_TOKEN_MAX - RTAS_TOKEN_BASE; token++) {
537 if (strcmp(cmd, rtas_table[token].name) == 0) {
538 SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
539 PowerPCCPU *cpu = POWERPC_CPU(first_cpu);
541 rtas_table[token].fn(cpu, spapr, token + RTAS_TOKEN_BASE,
542 nargs, args, nret, rets);
549 void spapr_rtas_register(int token, const char *name, spapr_rtas_fn fn)
551 assert((token >= RTAS_TOKEN_BASE) && (token < RTAS_TOKEN_MAX));
553 token -= RTAS_TOKEN_BASE;
555 assert(!name || !rtas_table[token].name);
557 rtas_table[token].name = name;
558 rtas_table[token].fn = fn;
561 void spapr_dt_rtas_tokens(void *fdt, int rtas)
565 for (i = 0; i < RTAS_TOKEN_MAX - RTAS_TOKEN_BASE; i++) {
566 struct rtas_call *call = &rtas_table[i];
572 _FDT(fdt_setprop_cell(fdt, rtas, call->name, i + RTAS_TOKEN_BASE));
576 hwaddr spapr_get_rtas_addr(void)
578 SpaprMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
580 const fdt32_t *rtas_data;
581 void *fdt = spapr->fdt_blob;
583 /* fetch rtas addr from fdt */
584 rtas_node = fdt_path_offset(fdt, "/rtas");
589 rtas_data = fdt_getprop(fdt, rtas_node, "linux,rtas-base", NULL);
595 * We assume that the OS called RTAS instantiate-rtas, but some other
596 * OS might call RTAS instantiate-rtas-64 instead. This fine as of now
597 * as SLOF only supports 32-bit variant.
599 return (hwaddr)fdt32_to_cpu(*rtas_data);
602 static void core_rtas_register_types(void)
604 spapr_rtas_register(RTAS_DISPLAY_CHARACTER, "display-character",
605 rtas_display_character);
606 spapr_rtas_register(RTAS_POWER_OFF, "power-off", rtas_power_off);
607 spapr_rtas_register(RTAS_SYSTEM_REBOOT, "system-reboot",
609 spapr_rtas_register(RTAS_QUERY_CPU_STOPPED_STATE, "query-cpu-stopped-state",
610 rtas_query_cpu_stopped_state);
611 spapr_rtas_register(RTAS_START_CPU, "start-cpu", rtas_start_cpu);
612 spapr_rtas_register(RTAS_STOP_SELF, "stop-self", rtas_stop_self);
613 spapr_rtas_register(RTAS_IBM_SUSPEND_ME, "ibm,suspend-me",
614 rtas_ibm_suspend_me);
615 spapr_rtas_register(RTAS_IBM_GET_SYSTEM_PARAMETER,
616 "ibm,get-system-parameter",
617 rtas_ibm_get_system_parameter);
618 spapr_rtas_register(RTAS_IBM_SET_SYSTEM_PARAMETER,
619 "ibm,set-system-parameter",
620 rtas_ibm_set_system_parameter);
621 spapr_rtas_register(RTAS_IBM_OS_TERM, "ibm,os-term",
623 spapr_rtas_register(RTAS_SET_POWER_LEVEL, "set-power-level",
624 rtas_set_power_level);
625 spapr_rtas_register(RTAS_GET_POWER_LEVEL, "get-power-level",
626 rtas_get_power_level);
627 spapr_rtas_register(RTAS_IBM_NMI_REGISTER, "ibm,nmi-register",
628 rtas_ibm_nmi_register);
629 spapr_rtas_register(RTAS_IBM_NMI_INTERLOCK, "ibm,nmi-interlock",
630 rtas_ibm_nmi_interlock);
633 type_init(core_rtas_register_types)