2 * PowerPC implementation of KVM hooks
4 * Copyright IBM Corp. 2007
7 * Jerone Young <jyoung5@us.ibm.com>
8 * Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
9 * Hollis Blanchard <hollisb@us.ibm.com>
11 * This work is licensed under the terms of the GNU GPL, version 2 or later.
12 * See the COPYING file in the top-level directory.
16 #include <sys/types.h>
17 #include <sys/ioctl.h>
20 #include <linux/kvm.h>
22 #include "qemu-common.h"
23 #include "qemu-timer.h"
28 #include "device_tree.h"
33 #define dprintf(fmt, ...) \
34 do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
36 #define dprintf(fmt, ...) \
40 /* XXX For some odd reason we sometimes hang inside KVM forever. I'd guess it's
41 * a race condition where we actually have a level triggered interrupt, but
42 * the infrastructure can't expose that yet, so the guest ACKs it, goes to
43 * sleep and never gets notified that there's still an interrupt pending.
45 * As a quick workaround, let's just wake up every 500 ms. That way we can
46 * assure that we're always reinjecting interrupts in time.
48 static QEMUTimer
*idle_timer
;
50 static void do_nothing(void *opaque
)
52 qemu_mod_timer(idle_timer
, qemu_get_clock(vm_clock
) +
53 (get_ticks_per_sec() / 2));
56 int kvm_arch_init(KVMState
*s
, int smp_cpus
)
61 int kvm_arch_init_vcpu(CPUState
*cenv
)
64 struct kvm_sregs sregs
;
66 sregs
.pvr
= cenv
->spr
[SPR_PVR
];
67 ret
= kvm_vcpu_ioctl(cenv
, KVM_SET_SREGS
, &sregs
);
72 void kvm_arch_reset_vcpu(CPUState
*env
)
76 int kvm_arch_put_registers(CPUState
*env
, int level
)
82 ret
= kvm_vcpu_ioctl(env
, KVM_GET_REGS
, ®s
);
92 regs
.srr0
= env
->spr
[SPR_SRR0
];
93 regs
.srr1
= env
->spr
[SPR_SRR1
];
95 regs
.sprg0
= env
->spr
[SPR_SPRG0
];
96 regs
.sprg1
= env
->spr
[SPR_SPRG1
];
97 regs
.sprg2
= env
->spr
[SPR_SPRG2
];
98 regs
.sprg3
= env
->spr
[SPR_SPRG3
];
99 regs
.sprg4
= env
->spr
[SPR_SPRG4
];
100 regs
.sprg5
= env
->spr
[SPR_SPRG5
];
101 regs
.sprg6
= env
->spr
[SPR_SPRG6
];
102 regs
.sprg7
= env
->spr
[SPR_SPRG7
];
104 for (i
= 0;i
< 32; i
++)
105 regs
.gpr
[i
] = env
->gpr
[i
];
107 ret
= kvm_vcpu_ioctl(env
, KVM_SET_REGS
, ®s
);
114 int kvm_arch_get_registers(CPUState
*env
)
116 struct kvm_regs regs
;
117 struct kvm_sregs sregs
;
120 ret
= kvm_vcpu_ioctl(env
, KVM_GET_REGS
, ®s
);
124 ret
= kvm_vcpu_ioctl(env
, KVM_GET_SREGS
, &sregs
);
134 env
->spr
[SPR_SRR0
] = regs
.srr0
;
135 env
->spr
[SPR_SRR1
] = regs
.srr1
;
137 env
->spr
[SPR_SPRG0
] = regs
.sprg0
;
138 env
->spr
[SPR_SPRG1
] = regs
.sprg1
;
139 env
->spr
[SPR_SPRG2
] = regs
.sprg2
;
140 env
->spr
[SPR_SPRG3
] = regs
.sprg3
;
141 env
->spr
[SPR_SPRG4
] = regs
.sprg4
;
142 env
->spr
[SPR_SPRG5
] = regs
.sprg5
;
143 env
->spr
[SPR_SPRG6
] = regs
.sprg6
;
144 env
->spr
[SPR_SPRG7
] = regs
.sprg7
;
146 for (i
= 0;i
< 32; i
++)
147 env
->gpr
[i
] = regs
.gpr
[i
];
149 #ifdef KVM_CAP_PPC_SEGSTATE
150 if (kvm_check_extension(env
->kvm_state
, KVM_CAP_PPC_SEGSTATE
)) {
151 env
->sdr1
= sregs
.u
.s
.sdr1
;
155 for (i
= 0; i
< 64; i
++) {
156 ppc_store_slb(env
, sregs
.u
.s
.ppc64
.slb
[i
].slbe
,
157 sregs
.u
.s
.ppc64
.slb
[i
].slbv
);
162 for (i
= 0; i
< 16; i
++) {
163 env
->sr
[i
] = sregs
.u
.s
.ppc32
.sr
[i
];
167 for (i
= 0; i
< 8; i
++) {
168 env
->DBAT
[0][i
] = sregs
.u
.s
.ppc32
.dbat
[i
] & 0xffffffff;
169 env
->DBAT
[1][i
] = sregs
.u
.s
.ppc32
.dbat
[i
] >> 32;
170 env
->IBAT
[0][i
] = sregs
.u
.s
.ppc32
.ibat
[i
] & 0xffffffff;
171 env
->IBAT
[1][i
] = sregs
.u
.s
.ppc32
.ibat
[i
] >> 32;
179 #if defined(TARGET_PPCEMB)
180 #define PPC_INPUT_INT PPC40x_INPUT_INT
181 #elif defined(TARGET_PPC64)
182 #define PPC_INPUT_INT PPC970_INPUT_INT
184 #define PPC_INPUT_INT PPC6xx_INPUT_INT
187 int kvm_arch_pre_run(CPUState
*env
, struct kvm_run
*run
)
193 idle_timer
= qemu_new_timer(vm_clock
, do_nothing
, NULL
);
194 qemu_mod_timer(idle_timer
, qemu_get_clock(vm_clock
) +
195 (get_ticks_per_sec() / 2));
198 /* PowerPC Qemu tracks the various core input pins (interrupt, critical
199 * interrupt, reset, etc) in PPC-specific env->irq_input_state. */
200 if (run
->ready_for_interrupt_injection
&&
201 (env
->interrupt_request
& CPU_INTERRUPT_HARD
) &&
202 (env
->irq_input_state
& (1<<PPC_INPUT_INT
)))
204 /* For now KVM disregards the 'irq' argument. However, in the
205 * future KVM could cache it in-kernel to avoid a heavyweight exit
206 * when reading the UIC.
210 dprintf("injected interrupt %d\n", irq
);
211 r
= kvm_vcpu_ioctl(env
, KVM_INTERRUPT
, &irq
);
213 printf("cpu %d fail inject %x\n", env
->cpu_index
, irq
);
216 /* We don't know if there are more interrupts pending after this. However,
217 * the guest will return to userspace in the course of handling this one
218 * anyways, so we will get a chance to deliver the rest. */
222 int kvm_arch_post_run(CPUState
*env
, struct kvm_run
*run
)
227 static int kvmppc_handle_halt(CPUState
*env
)
229 if (!(env
->interrupt_request
& CPU_INTERRUPT_HARD
) && (msr_ee
)) {
231 env
->exception_index
= EXCP_HLT
;
237 /* map dcr access to existing qemu dcr emulation */
238 static int kvmppc_handle_dcr_read(CPUState
*env
, uint32_t dcrn
, uint32_t *data
)
240 if (ppc_dcr_read(env
->dcr_env
, dcrn
, data
) < 0)
241 fprintf(stderr
, "Read to unhandled DCR (0x%x)\n", dcrn
);
246 static int kvmppc_handle_dcr_write(CPUState
*env
, uint32_t dcrn
, uint32_t data
)
248 if (ppc_dcr_write(env
->dcr_env
, dcrn
, data
) < 0)
249 fprintf(stderr
, "Write to unhandled DCR (0x%x)\n", dcrn
);
254 int kvm_arch_handle_exit(CPUState
*env
, struct kvm_run
*run
)
258 switch (run
->exit_reason
) {
260 if (run
->dcr
.is_write
) {
261 dprintf("handle dcr write\n");
262 ret
= kvmppc_handle_dcr_write(env
, run
->dcr
.dcrn
, run
->dcr
.data
);
264 dprintf("handle dcr read\n");
265 ret
= kvmppc_handle_dcr_read(env
, run
->dcr
.dcrn
, &run
->dcr
.data
);
269 dprintf("handle halt\n");
270 ret
= kvmppc_handle_halt(env
);
277 static int read_cpuinfo(const char *field
, char *value
, int len
)
281 int field_len
= strlen(field
);
284 f
= fopen("/proc/cpuinfo", "r");
290 if(!fgets(line
, sizeof(line
), f
)) {
293 if (!strncmp(line
, field
, field_len
)) {
294 strncpy(value
, line
, len
);
305 uint32_t kvmppc_get_tbfreq(void)
309 uint32_t retval
= get_ticks_per_sec();
311 if (read_cpuinfo("timebase", line
, sizeof(line
))) {
315 if (!(ns
= strchr(line
, ':'))) {