Skip to contentSkip to conversation

KVM: x86: remove duplicated KVM_REQ_EVENT request

PATCH linmiaohe 3 participants 3 messages Feb 7, 2020 – Feb 12, 2020 KVM

lore
Share & tools
lore

linmiaohe

author
#0 lore raw

Hi:

Quoted 12 lines · Vitaly Kuznetsov

Review

Quoted 7 lines · linmiaohe

From: Miaohe Lin <[email protected]>

The KVM_REQ_EVENT request is already made in kvm_set_rflags(). We should not make it again.

(quoted diff) +0 -1
 	kvm_rip_write(vcpu, ctxt->eip);
 	kvm_set_rflags(vcpu, ctxt->eflags);
-	kvm_make_request(KVM_REQ_EVENT, vcpu);

I would've actually done it the other way around and removed
kvm_make_request() from kvm_set_rflags() as it is not an obvious behavior (e.g. why kvm_rip_write() doens't do that and kvm_set_rflags() does ?) adding kvm_make_request() to all call sites.

Quoted 3 lines

In case this looks like too big of a change with no particular gain I'd suggest you at least leave a comment above kvm_set_rflags(), something like

"kvm_make_request() also requests KVM_REQ_EVENT"

I think adding kvm_make_request() to all call sites is too big without particular gain. And also leave a comment above kvm_set_rflags() maybe isn't needed as rflags updates is an site that can trigger event injection. Please see commit (3842d135ff24 KVM: Check for pending events before attempting injection) for detail.

What do you think? Thanks for your review anyway. :)

Reviewed-by
#1 lore raw
Quoted 16 lines · linmiaohe

Hi:

Quoted 14 lines · Vitaly Kuznetsov

Review

Quoted 7 lines · linmiaohe

From: Miaohe Lin <[email protected]>

The KVM_REQ_EVENT request is already made in kvm_set_rflags(). We should not make it again.

(quoted diff) +0 -1
 	kvm_rip_write(vcpu, ctxt->eip);
 	kvm_set_rflags(vcpu, ctxt->eflags);
-	kvm_make_request(KVM_REQ_EVENT, vcpu);

I would've actually done it the other way around and removed
kvm_make_request() from kvm_set_rflags() as it is not an obvious behavior (e.g. why kvm_rip_write() doens't do that and kvm_set_rflags() does ?) adding kvm_make_request() to all call sites.

In case this looks like too big of a change with no particular gain I'd suggest you at least leave a comment above kvm_set_rflags(), something like

Quoted 7 lines
Quoted 1 line

"kvm_make_request() also requests KVM_REQ_EVENT"

I think adding kvm_make_request() to all call sites is too big without particular gain. And also leave a comment above kvm_set_rflags() maybe isn't needed as rflags updates is an site that can trigger event injection. Please see commit (3842d135ff24 KVM: Check for pending events before attempting injection) for detail.

What do you think?

I don't have a strong opinion on this and your change is correct so feel free to throw my

Signature · 1 line
Vitaly
maintainer
#2 lore raw
Quoted 3 lines · Vitaly Kuznetsov

kvm_make_request() from kvm_set_rflags() as it is not an obvious behavior (e.g. why kvm_rip_write() doens't do that and kvm_set_rflags() does ?)

Because writing RFLAGS can change IF and therefore cause an interrupt to be injected.

Paolo