i386/kvm: Cleanups in kvm_arch_init()
cover i386/kvm: Cleanups in kvm_arch_init()
Patch 1 removes the object_dynamic_cast() check in kvm_arch_init();
Patch 2 removes the unncessary kvm_check_extension(s, KVM_CAP_X86_SMM);
Xiaoyao Li (2):
i386/kvm: Get X86MachineState in kvm_arch_init() without the cast
check
i386/kvm: Drop KVM_CAP_X86_SMM check in kvm_arch_init()
target/i386/kvm/kvm.c | 23 +++++++++-------------- 1 file changed, 9 insertions(+), 14 deletions(-)
Signature · 1 line
2.43.0
PATCH 1/2 i386/kvm: Get X86MachineState in kvm_arch_init() without the cast check
Commit 8f54bbd0b4d9 ("x86: Check for machine state object class before typecasting it") added back the object_dynamic_cast() check before casting MachineState to X86MachineState. And commit 035d1ef26565 ("i386:
Add ratelimit for bus locks acquired in guest") followed it.
The reason to check object_dynamic_cast(OBJECT(ms), TYPE_PC_MACHINE) before commit 8f54bbd0b4d9 was that smm was not supported for microvm machine at that time. But after commit 8f54bbd0b4d9, smm is supported for all x86 machines (both pc and microvm). And since it's the target-specifc implementation of kvm_arch_init() in target/i386/kvm/kvm.c, I don't see how it would be called for other machines than x86machine, and why the check of object_dynamic_cast() is needed.
Drop the object_dynamic_cast() check and simplify the code.
--- target/i386/kvm/kvm.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-)
target/i386/kvm/kvm.c +9 -13
@@ -3230,6 +3230,7 @@ static int kvm_vm_enable_energy_msrs(KVMState *s)int kvm_arch_init (MachineState *ms,KVMState *s) { +X86MachineState *x86ms =X86_MACHINE (ms);int ret;struct utsname utsname;Error *local_err =NULL ;
@@ -3311,8 +3312,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s) }if (kvm_check_extension (s,KVM_CAP_X86_SMM ) && -object_dynamic_cast (OBJECT (ms),TYPE_X86_MACHINE ) && -x86_machine_is_smm_enabled (X86_MACHINE (ms))) { +x86_machine_is_smm_enabled (x86ms)) { smram_machine_done.notify = register_smram_listener;qemu_add_machine_init_done_notifier (&smram_machine_done); }
@@ -3326,18 +3326,14 @@ int kvm_arch_init(MachineState *ms, KVMState *s) } } -if (object_dynamic_cast (OBJECT (ms),TYPE_X86_MACHINE )) { -X86MachineState *x86ms =X86_MACHINE (ms); - -if (x86ms->bus_lock_ratelimit >0 ) { - ret =kvm_vm_enable_bus_lock_exit (s); -if (ret <0 ) { -return ret; - } -ratelimit_init (&bus_lock_ratelimit_ctrl); -ratelimit_set_speed (&bus_lock_ratelimit_ctrl, - x86ms->bus_lock_ratelimit,BUS_LOCK_SLICE_TIME ); +if (x86ms->bus_lock_ratelimit >0 ) { + ret =kvm_vm_enable_bus_lock_exit (s); +if (ret <0 ) { +return ret; } +ratelimit_init (&bus_lock_ratelimit_ctrl); +ratelimit_set_speed (&bus_lock_ratelimit_ctrl, + x86ms->bus_lock_ratelimit,BUS_LOCK_SLICE_TIME ); }if (kvm_check_extension (s,KVM_CAP_X86_NOTIFY_VMEXIT )) {
Signature · 1 line
2.43.0
Re: [PATCH 1/2] i386/kvm: Get X86MachineState in kvm_arch_init() without the cast check
Quoted 13 lines · Xiaoyao Li
Commit
8f54bbd0b4d9("x86: Check for machine state object class before typecasting it") added back theobject_dynamic_cast()check before casting MachineState to X86MachineState. And commit035d1ef26565("i386:
Add ratelimit for bus locks acquired in guest") followed it.The reason to check
object_dynamic_cast(OBJECT(ms), TYPE_PC_MACHINE)before commit8f54bbd0b4d9was that smm was not supported for microvm machine at that time. But after commit8f54bbd0b4d9, smm is supported for all x86 machines (both pc and microvm). And since it's the target-specifc implementation ofkvm_arch_init()intarget/i386/kvm/kvm.c, I don't see how it would be called for other machines than x86machine, and why the check ofobject_dynamic_cast()is needed.
Quoted 6 lines
Drop the
object_dynamic_cast()check and simplify the code.Signed-off-by Xiaoyao Li <[email protected]>--- target/i386/kvm/kvm.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-)
Re: [PATCH 1/2] i386/kvm: Get X86MachineState in kvm_arch_init() without the cast check
In reply to Xiaoyao LiQuoted 13 lines · Xiaoyao Li
Commit
8f54bbd0b4d9("x86: Check for machine state object class before typecasting it") added back theobject_dynamic_cast()check before casting MachineState to X86MachineState. And commit035d1ef26565("i386:
Add ratelimit for bus locks acquired in guest") followed it.The reason to check
object_dynamic_cast(OBJECT(ms), TYPE_PC_MACHINE)before commit8f54bbd0b4d9was that smm was not supported for microvm machine at that time. But after commit8f54bbd0b4d9, smm is supported for all x86 machines (both pc and microvm). And since it's the target-specifc implementation ofkvm_arch_init()intarget/i386/kvm/kvm.c, I don't see how it would be called for other machines than x86machine, and why the check ofobject_dynamic_cast()is needed.
Quoted 3 lines
Drop the
object_dynamic_cast()check and simplify the code.Signed-off-by Xiaoyao Li <[email protected]>
LGTM.
Quoted 52 lines
--- target/i386/kvm/kvm.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-)target/i386/kvm/kvm.c +9
-13@@ -3230,6 +3230,7 @@ static int kvm_vm_enable_energy_msrs(KVMState *s)int kvm_arch_init (MachineState *ms,KVMState *s) { +X86MachineState *x86ms =X86_MACHINE (ms);int ret;struct utsname utsname;Error *local_err =NULL ;@@ -3311,8 +3312,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s) }if (kvm_check_extension (s,KVM_CAP_X86_SMM ) && -object_dynamic_cast (OBJECT (ms),TYPE_X86_MACHINE ) && -x86_machine_is_smm_enabled (X86_MACHINE (ms))) { +x86_machine_is_smm_enabled (x86ms)) { smram_machine_done.notify = register_smram_listener;qemu_add_machine_init_done_notifier (&smram_machine_done); }@@ -3326,18 +3326,14 @@ int kvm_arch_init(MachineState *ms, KVMState *s) } } -if (object_dynamic_cast (OBJECT (ms),TYPE_X86_MACHINE )) { -X86MachineState *x86ms =X86_MACHINE (ms); - -if (x86ms->bus_lock_ratelimit >0 ) { - ret =kvm_vm_enable_bus_lock_exit (s); -if (ret <0 ) { -return ret; - } -ratelimit_init (&bus_lock_ratelimit_ctrl); -ratelimit_set_speed (&bus_lock_ratelimit_ctrl, - x86ms->bus_lock_ratelimit,BUS_LOCK_SLICE_TIME ); +if (x86ms->bus_lock_ratelimit >0 ) { + ret =kvm_vm_enable_bus_lock_exit (s); +if (ret <0 ) { +return ret; } +ratelimit_init (&bus_lock_ratelimit_ctrl); +ratelimit_set_speed (&bus_lock_ratelimit_ctrl, + x86ms->bus_lock_ratelimit,BUS_LOCK_SLICE_TIME ); }if (kvm_check_extension (s,KVM_CAP_X86_NOTIFY_VMEXIT )) {
PATCH 2/2 i386/kvm: Drop KVM_CAP_X86_SMM check in kvm_arch_init()
In reply to Xiaoyao Lix86_machine_is_smm_enabled() checks the KVM_CAP_X86_SMM for KVM case. No need to check KVM_CAP_X86_SMM specifically.
One benefit of checking it specifically is that it can return early when KVM_CAP_X86_SMM is not supported. But considering KVM_CAP_X86_SMM was added in KVM 10 year ago, it's unlikely the case.
So just drop the check of KVM_CAP_X86_SMM to simplify the code.
--- target/i386/kvm/kvm.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
target/i386/kvm/kvm.c +1 -2
@@ -3311,8 +3311,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s)return ret; } -if (kvm_check_extension (s,KVM_CAP_X86_SMM ) && -x86_machine_is_smm_enabled (x86ms)) { +if (x86_machine_is_smm_enabled (x86ms)) { smram_machine_done.notify = register_smram_listener;qemu_add_machine_init_done_notifier (&smram_machine_done); }
Signature · 1 line
2.43.0
Quoted 3 lines · Xiaoyao Li
Patch 1 removes the
object_dynamic_cast()check inkvm_arch_init();Patch 2 removes the unncessary
kvm_check_extension(s, KVM_CAP_X86_SMM);
Gentle ping.
Quoted 8 lines
Xiaoyao Li (2):
i386/kvm: Get X86MachineState inkvm_arch_init()without the cast
check
i386/kvm: DropKVM_CAP_X86_SMMcheck inkvm_arch_init()target/i386/kvm/kvm.c | 23 +++++++++-------------- 1 file changed, 9 insertions(+), 14 deletions(-)