On Mon, Nov 27, 2017 at 05:53:01PM +0100, Paolo Bonzini wrote:
> On 25/11/2017 21:57, Christoffer Dall wrote:
> > In preparation for moving calls to vcpu_load() and vcpu_put() into the
> > architecture specific implementations of the KVM vcpu ioctls, move the
> > calls in the main kvm_vcpu_ioctl() dispatcher function to each case
> > of the ioctl select statement. This allows us to move the vcpu_load()
> > and vcpu_put() calls into architecture specific implementations of vcpu
> > ioctls, one by one.
> >
> > Signed-off-by: Christoffer Dall <christoffer.dall@xxxxxxxxxx>
> > ---
> > virt/kvm/kvm_main.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++-----
> > 1 file changed, 48 insertions(+), 5 deletions(-)
> >
> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index 9deb5a245b83..fafafcc38b5a 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -2528,16 +2528,15 @@ static long kvm_vcpu_ioctl(struct file *filp,
> > return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
> > #endif
> >
> > -
> > - r = vcpu_load(vcpu);
> > - if (r)
> > - return r;
> > switch (ioctl) {
> > case KVM_RUN: {
> > struct pid *oldpid;
> > r = -EINVAL;
> > if (arg)
> > goto out;
> > + r = vcpu_load(vcpu);
> > + if (r)
> > + goto out;
> > oldpid = rcu_access_pointer(vcpu->pid);
>
> If it is not a problem for ARM, maybe it would actually be best to leave
> the locking in kvm_vcpu_ioctl (with the already existing exception of
> KVM_INTERRUPT). This would make vcpu_load void, and would also let you
> keep the PID adjustment in common code. This would be more similar to
> the previous version, but without introducing __vcpu_load/__vcpu_put.
Yes, that's not a problem for ARM, and it was actually what I started
out with, and you can see the result here (rebased on v4.15-rc1):
git://git.kernel.org/pub/scm/linux/kernel/git/cdall/linux.git vcpu-load-put-keeplock
I got a bit into getting rid of the (IMHO) ugly ifdef-shortcut
dispatcher code, and thus reworked it to the submitted version.
Going back and looking, it's nicer to avoid the pid adjustment call, and
having vcpu_load be void is also convenient, but we're stuck with the
ifdef. I guess I lean towards your suggestion as well, given that my
problem with the ifdef is not a technical one, but an aesthetic one.
>
> Looks good apart from this doubt! Thanks,
>
Let me know if you want to have a quick glance at the branch above and
prefer that I send that as v2.
Thanks,
-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe linux-s390" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
