Re: [PATCH v3 2/6] KVM: x86: add force_intercept_exceptions_mask

[Sean Christopherson <seanjc@google.com>]

On Tue, Feb 08, 2022, Maxim Levitsky wrote:
> On Thu, 2021-09-02 at 16:56 +0000, Sean Christopherson wrote:
> > Assuming this hasn't been abandoned...
> > 
> > On Wed, Aug 11, 2021, Maxim Levitsky wrote:
> > > This parameter will be used by VMX and SVM code to force
> > > interception of a set of exceptions, given by a bitmask
> > > for guest debug and/or kvm debug.
> > > 
> > > This is based on an idea first shown here:
> > > https://patchwork.kernel.org/project/kvm/patch/20160301192822.GD22677@pd.tnic/
> > > 
> > > CC: Borislav Petkov <bp@suse.de>
> > > Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> > > ---
> > >  arch/x86/kvm/x86.c | 3 +++
> > >  arch/x86/kvm/x86.h | 2 ++
> > >  2 files changed, 5 insertions(+)
> > > 
> > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > > index fdc0c18339fb..092e2fad3c0d 100644
> > > --- a/arch/x86/kvm/x86.c
> > > +++ b/arch/x86/kvm/x86.c
> > > @@ -184,6 +184,9 @@ module_param(force_emulation_prefix, bool, S_IRUGO);
> > >  int __read_mostly pi_inject_timer = -1;
> > >  module_param(pi_inject_timer, bint, S_IRUGO | S_IWUSR);
> > >  
> > > +uint force_intercept_exceptions_mask;
> > > +module_param(force_intercept_exceptions_mask, uint, S_IRUGO | S_IWUSR);
> > 
> > Use octal permissions.  This also can't be a simple writable param, at least not
> > without a well-documented disclaimer, as there's no guarantee a vCPU will update
> > its exception bitmap in a timely fashion.  An alternative to a module param would
> > be to extend/add a per-VM ioctl(), e.g. maybe KVM_SET_GUEST_DEBUG?  The downside
> > of an ioctl() is that it would require userspace enabling :-/
> > 
> 
> All other module params in this file use macros for permissions, that is why
> I used them too.
> 
> I'll add a comment with a disclaimer here - this is only for debug.
> I strongly don't want to have this as ioctl as that will indeed need qemu patches,
> not to mention things like unit tests and which don't even always use qemu.
> 
> Or I can make this parameter read-only. I don't mind reloading kvm module when
> I change this parameter.

Oh!  We can force an update, a la nx_huge_pages, where the setter loops through
all VMs and does a kvm_make_all_cpus_request() to instruct vCPUs to update their
bitmaps.  Requires a new request, but that doesn't seem like a huge deal, and it
might help pave the way for adding more debug hooks for developers.

The param should also be "unsafe".

E.g. something like

static const struct kernel_param_ops force_ex_intercepts_ops = {
	.set = set_force_exception_intercepts,
	.get = get_force_exception_intercepts,
};
module_param_cb_unsafe(force_exception_intercepts, &force_ex_intercepts_ops,
		       &force_exception_intercepts, 0644);