Handle INIT before SIPI.

PATCH Applied Gleb Natapov 3 participants 4 messages Oct 16, 2008 – Oct 22, 2008 KVM

lore

Share & tools

lore

Avi Kivity

Applied

2008-10-19 09:42 UTC #1 lore raw

Quoted 5 lines · Gleb Natapov

Handle INIT before SIPI.
If SIPI was issued before CPU handled INIT signal the CPU is not reseted.

Applied, thanks.

Signature · 1 line

error compiling committee.c: too many arguments to function

Xu, Dongxiao

2008-10-21 02:34 UTC #2 lore raw

In reply to Gleb Natapov

Hi, Gleb,
It seems that this patch is x86 specific, and it breaks the IA-64 compilation. Do you have any good idea to solve this? Thanks!

Best Regards,
-- Dongxiao

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Gleb Natapov
Sent: 2008年10月16日 23:21
To: [email protected]
Cc: [email protected]
Subject: Handle INIT before SIPI.

Handle INIT before SIPI.

If SIPI was issued before CPU handled INIT signal the CPU is not reseted.

Signed-off-by Gleb Natapov <[email protected]>

qemu/qemu-kvm.c +14 -5

index 577f2b2..dae05b0 100644
@@ -349,13 +349,20 @@ static void update_regs_for_sipi(CPUState *env)
 {
     kvm_arch_update_regs_for_sipi(env);
     vcpu_info[env->cpu_index].sipi_needed = 0;
-    vcpu_info[env->cpu_index].init = 0;
 }
 
 static void update_regs_for_init(CPUState *env)
 {
+    SegmentCache cs = env->segs[R_CS];
+
     cpu_reset(env);
+
+    /* restore SIPI vector */
+    if(vcpu_info[env->cpu_index].sipi_needed)
+        env->segs[R_CS] = cs;
+
     kvm_arch_load_regs(env);
+    vcpu_info[env->cpu_index].init = 0;
 }
 
 static void setup_kernel_sigmask(CPUState *env)

@@ -411,10 +418,12 @@ static int kvm_main_loop_cpu(CPUState *env)
            kvm_main_loop_wait(env, 1000);
        if (env->interrupt_request & (CPU_INTERRUPT_HARD | CPU_INTERRUPT_NMI))
            env->halted = 0;
-       if (!kvm_irqchip_in_kernel(kvm_context) && info->sipi_needed)
-           update_regs_for_sipi(env);
-       if (!kvm_irqchip_in_kernel(kvm_context) && info->init)
-           update_regs_for_init(env);
+    if (!kvm_irqchip_in_kernel(kvm_context)) {
+           if (info->init)
+               update_regs_for_init(env);
+           if (info->sipi_needed)
+               update_regs_for_sipi(env);
+    }
        if (!env->halted && !info->init)
            kvm_cpu_exec(env);
        env->interrupt_request &= ~CPU_INTERRUPT_EXIT;

--
Gleb.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Gleb Natapov

author

2008-10-22 12:48 UTC #3 lore raw

Quoted 3 lines · Dongxiao

Hi, Gleb,
It seems that this patch is x86 specific, and it breaks the IA-64 compilation. Do you have any good idea to solve this? Thanks!

Solve this by ifdefs for now. For better solution cleanup of SMP handling is needed.

Signed-off-by Gleb Natapov <[email protected]>

qemu/qemu-kvm.c +5 -1

index 67d1de9..c5f3f29 100644
@@ -353,16 +353,20 @@ static void update_regs_for_sipi(CPUState *env)
 
 static void update_regs_for_init(CPUState *env)
 {
+#ifdef TARGET_I386
     SegmentCache cs = env->segs[R_CS];
+#endif
 
     cpu_reset(env);
 
+#ifdef TARGET_I386
     /* restore SIPI vector */
     if(vcpu_info[env->cpu_index].sipi_needed)
         env->segs[R_CS] = cs;
 
-    kvm_arch_load_regs(env);
     vcpu_info[env->cpu_index].init = 0;
+#endif
+    kvm_arch_load_regs(env);
 }
 
 static void setup_kernel_sigmask(CPUState *env)

--
Gleb.