[PATCH 3/3 v3] Enable kvm emulated watchdog

[PATCH 3/3 v3] Enable kvm emulated watchdog

[Bharat Bhushan]

Enable the KVM emulated watchdog if KVM supports (use the
capability enablement in watchdog handler). Also watchdog exit
(KVM_EXIT_WATCHDOG) handling is added.
Watchdog state machine is cleared whenever VM state changes to running.
This is to handle the cases like return from debug halt etc.

Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
---
v3:
 - TSR clearing is removed in whatchdog exit handling as this is no more needed.

v2:
 - Merged ([PATCH 3/4] Watchdog exit handling support) and ([PATCH 4/4] Enable to use kvm emulated watchdog)
 - Clear watchdog state machine when VM state changes to running.

 hw/ppc_booke.c            |    5 +++
 linux-headers/linux/kvm.h |    2 +
 target-ppc/cpu.h          |    1 +
 target-ppc/kvm.c          |   63 +++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 71 insertions(+), 0 deletions(-)

diff --git a/hw/ppc_booke.c b/hw/ppc_booke.c
index 837a5b6..a9fba15 100644
--- a/hw/ppc_booke.c
+++ b/hw/ppc_booke.c
@@ -203,6 +203,11 @@ static void booke_wdt_cb(void *opaque)
                              booke_timer->wdt_timer);
 }
 
+void ppc_booke_wdt_clear_tsr(CPUPPCState *env, target_ulong tsr)
+{
+    env->spr[SPR_BOOKE_TSR] = tsr & ~(TSR_ENW | TSR_WIS | TSR_WRS_MASK);
+}
+
 void store_booke_tsr(CPUPPCState *env, target_ulong val)
 {
     env->spr[SPR_BOOKE_TSR] &= ~val;
diff --git a/linux-headers/linux/kvm.h b/linux-headers/linux/kvm.h
index 4b9e575..64d94da 100644
--- a/linux-headers/linux/kvm.h
+++ b/linux-headers/linux/kvm.h
@@ -163,6 +163,7 @@ struct kvm_pit_config {
 #define KVM_EXIT_OSI              18
 #define KVM_EXIT_PAPR_HCALL	  19
 #define KVM_EXIT_S390_UCONTROL	  20
+#define KVM_EXIT_WATCHDOG         21
 
 /* For KVM_EXIT_INTERNAL_ERROR */
 #define KVM_INTERNAL_ERROR_EMULATION 1
@@ -618,6 +619,7 @@ struct kvm_ppc_smmu_info {
 #define KVM_CAP_PPC_GET_SMMU_INFO 78
 #define KVM_CAP_S390_COW 79
 #define KVM_CAP_PPC_ALLOC_HTAB 80
+#define KVM_CAP_PPC_WDT	81
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
diff --git a/target-ppc/cpu.h b/target-ppc/cpu.h
index ca2fc21..78212b4 100644
--- a/target-ppc/cpu.h
+++ b/target-ppc/cpu.h
@@ -1191,6 +1191,7 @@ void store_40x_dbcr0 (CPUPPCState *env, uint32_t val);
 void store_40x_sler (CPUPPCState *env, uint32_t val);
 void store_booke_tcr (CPUPPCState *env, target_ulong val);
 void store_booke_tsr (CPUPPCState *env, target_ulong val);
+void ppc_booke_wdt_clear_tsr(CPUPPCState *env, target_ulong tsr);
 void booke206_flush_tlb(CPUPPCState *env, int flags, const int check_iprot);
 target_phys_addr_t booke206_tlb_to_page_size(CPUPPCState *env, ppcmas_tlb_t *tlb);
 int ppcmas_tlb_check(CPUPPCState *env, ppcmas_tlb_t *tlb,
diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
index b6ef72d..393214e 100644
--- a/target-ppc/kvm.c
+++ b/target-ppc/kvm.c
@@ -32,6 +32,7 @@
 #include "device_tree.h"
 #include "hw/sysbus.h"
 #include "hw/spapr.h"
+#include "hw/watchdog.h"
 
 #include "hw/sysbus.h"
 #include "hw/spapr.h"
@@ -60,6 +61,7 @@ static int cap_booke_sregs;
 static int cap_ppc_smt;
 static int cap_ppc_rma;
 static int cap_spapr_tce;
+static int cap_ppc_wdt;
 
 /* XXX We have a race condition where we actually have a level triggered
  *     interrupt, but the infrastructure can't expose that yet, so the guest
@@ -86,6 +88,7 @@ int kvm_arch_init(KVMState *s)
     cap_ppc_smt = kvm_check_extension(s, KVM_CAP_PPC_SMT);
     cap_ppc_rma = kvm_check_extension(s, KVM_CAP_PPC_RMA);
     cap_spapr_tce = kvm_check_extension(s, KVM_CAP_SPAPR_TCE);
+    cap_ppc_wdt = kvm_check_extension(s, KVM_CAP_PPC_WDT);
 
     if (!cap_interrupt_level) {
         fprintf(stderr, "KVM: Couldn't find level irq capability. Expect the "
@@ -168,6 +171,25 @@ static int kvm_booke206_tlb_init(CPUPPCState *env)
     return 0;
 }
 
+static int kvm_wdt_enable(CPUPPCState *env)
+{
+    int ret;
+    struct kvm_enable_cap encap = {};
+
+    if (!kvm_enabled() || !cap_ppc_wdt) {
+        return 0;
+    }
+
+    encap.cap = KVM_CAP_PPC_WDT;
+    ret = kvm_vcpu_ioctl(env, KVM_ENABLE_CAP, &encap);
+    if (ret < 0) {
+        fprintf(stderr, "%s: couldn't enable KVM_CAP_PPC_WDT: %s\n",
+                __func__, strerror(-ret));
+        return ret;
+    }
+
+    return ret;
+}
 
 #if defined(TARGET_PPC64)
 static void kvm_get_fallback_smmu_info(CPUPPCState *env,
@@ -371,6 +393,33 @@ static inline void kvm_fixup_page_sizes(CPUPPCState *env)
 
 #endif /* !defined (TARGET_PPC64) */
 
+static void cpu_state_change_handler(void *opaque, int running, RunState state)
+{
+    CPUPPCState *env = opaque;
+
+    struct kvm_sregs sregs;
+
+    printf("running = %d, state = %d \n", running, state);
+    if (!running)
+        return;
+
+    /*
+     * Clear watchdog interrupt condition by clearing TSR.
+     * Similar logic needed to be implemented for watchdog
+     * emulation in qemu.
+     */
+    if (cap_booke_sregs && cap_ppc_wdt) {
+        kvm_vcpu_ioctl(env, KVM_GET_SREGS, &sregs);
+
+        /* Clear TSR.ENW, TSR.WIS and TSR.WRS */
+        ppc_booke_wdt_clear_tsr(env, sregs.u.e.tsr);
+        sregs.u.e.tsr = env->spr[SPR_BOOKE_TSR];
+        sregs.u.e.update_special = KVM_SREGS_E_BASE | KVM_SREGS_E_UPDATE_TSR;
+
+        kvm_vcpu_ioctl(env, KVM_SET_SREGS, &sregs);
+    }
+}
+
 int kvm_arch_init_vcpu(CPUPPCState *cenv)
 {
     int ret;
@@ -384,6 +433,13 @@ int kvm_arch_init_vcpu(CPUPPCState *cenv)
         return ret;
     }
 
+    ret = kvm_wdt_enable(cenv);
+    if (ret) {
+        return ret;
+    }
+
+    qemu_add_vm_change_state_handler(cpu_state_change_handler, cenv);
+
     idle_timer = qemu_new_timer_ns(vm_clock, kvm_kick_env, cenv);
 
     /* Some targets support access to KVM's guest TLB. */
@@ -769,6 +825,13 @@ int kvm_arch_handle_exit(CPUPPCState *env, struct kvm_run *run)
         ret = 1;
         break;
 #endif
+#ifdef KVM_EXIT_WATCHDOG
+    case KVM_EXIT_WATCHDOG:
+        dprintf("booke watchdog action\n");
+    	watchdog_perform_action();
+        ret = 0;
+        break;
+#endif
     default:
         fprintf(stderr, "KVM: unknown exit reason %d\n", run->exit_reason);
         ret = -1;
-- 
1.7.0.4

Re: [PATCH 3/3 v3] Enable kvm emulated watchdog

[Scott Wood]

On 07/18/2012 06:34 AM, Bharat Bhushan wrote:
> diff --git a/linux-headers/linux/kvm.h b/linux-headers/linux/kvm.h
> index 4b9e575..64d94da 100644
> --- a/linux-headers/linux/kvm.h
> +++ b/linux-headers/linux/kvm.h
> @@ -163,6 +163,7 @@ struct kvm_pit_config {
>  #define KVM_EXIT_OSI              18
>  #define KVM_EXIT_PAPR_HCALL	  19
>  #define KVM_EXIT_S390_UCONTROL	  20
> +#define KVM_EXIT_WATCHDOG         21
>  
>  /* For KVM_EXIT_INTERNAL_ERROR */
>  #define KVM_INTERNAL_ERROR_EMULATION 1
> @@ -618,6 +619,7 @@ struct kvm_ppc_smmu_info {
>  #define KVM_CAP_PPC_GET_SMMU_INFO 78
>  #define KVM_CAP_S390_COW 79
>  #define KVM_CAP_PPC_ALLOC_HTAB 80
> +#define KVM_CAP_PPC_WDT	81
>  
>  #ifdef KVM_CAP_IRQ_ROUTING
>  

linux headers update should be a separate patch.

> @@ -371,6 +393,33 @@ static inline void kvm_fixup_page_sizes(CPUPPCState *env)
>  
>  #endif /* !defined (TARGET_PPC64) */
>  
> +static void cpu_state_change_handler(void *opaque, int running, RunState state)
> +{
> +    CPUPPCState *env = opaque;
> +
> +    struct kvm_sregs sregs;
> +
> +    printf("running = %d, state = %d \n", running, state);

Remove this printf.

> +    if (!running)
> +        return;
> +
> +    /*
> +     * Clear watchdog interrupt condition by clearing TSR.
> +     * Similar logic needed to be implemented for watchdog
> +     * emulation in qemu.
> +     */
> +    if (cap_booke_sregs && cap_ppc_wdt) {
> +        kvm_vcpu_ioctl(env, KVM_GET_SREGS, &sregs);
> +
> +        /* Clear TSR.ENW, TSR.WIS and TSR.WRS */
> +        ppc_booke_wdt_clear_tsr(env, sregs.u.e.tsr);
> +        sregs.u.e.tsr = env->spr[SPR_BOOKE_TSR];
> +        sregs.u.e.update_special = KVM_SREGS_E_BASE | KVM_SREGS_E_UPDATE_TSR;

KVM_SREGS_E_BASE is for "features", not "update_special".  Used here
it'll translate to KVM_SREGS_E_UPDATE_MCSR.

> +        kvm_vcpu_ioctl(env, KVM_SET_SREGS, &sregs);
> +    }
> +}
> +
>  int kvm_arch_init_vcpu(CPUPPCState *cenv)
>  {
>      int ret;
> @@ -384,6 +433,13 @@ int kvm_arch_init_vcpu(CPUPPCState *cenv)
>          return ret;
>      }
>  
> +    ret = kvm_wdt_enable(cenv);
> +    if (ret) {
> +        return ret;
> +    }

s/wdt/watchdog/g

Only do this if the user asks for a watchdog enforcement action -- and
if they do, make sure to at least warn if KVM doesn't support it.

> +    qemu_add_vm_change_state_handler(cpu_state_change_handler, cenv);
> +
>      idle_timer = qemu_new_timer_ns(vm_clock, kvm_kick_env, cenv);
>  
>      /* Some targets support access to KVM's guest TLB. */
> @@ -769,6 +825,13 @@ int kvm_arch_handle_exit(CPUPPCState *env, struct kvm_run *run)
>          ret = 1;
>          break;
>  #endif
> +#ifdef KVM_EXIT_WATCHDOG
> +    case KVM_EXIT_WATCHDOG:
> +        dprintf("booke watchdog action\n");
> +    	watchdog_perform_action();
> +        ret = 0;
> +        break;

Indentation.  Is this part really booke-specific?

-Scott