ACK/Cmnt: [SRU][Xenial][PATCH v2 0/3] Fix corruption on blocked_vcpu_on_cpu list

[Ioanna Alifieraki]

On Wed, Jan 20, 2021 at 8:51 AM Stefan Bader <stefan.bader at canonical.com> wrote:
>
> On 06.01.21 00:54, Ioanna Alifieraki wrote:
> > BugLink: https://bugs.launchpad.net/bugs/1908428
> >
> > The following 3 patches fix the bug reported in [1].
> > They are backported to apply on 4.4 kernels.
> > Backport is required becasue upstream commit
> > bc22512bb24c(kvm: vmx: rename vmx_pre/post_block to pi_pre/post_block)
> > which renames  vmx_pre/post_block functions to pi_pre/post_block
> > is missing from 4.4.
> > Original patches come from [2] and have been accepted upstream.
> >
> > [1] https://marc.info/?l=kvm&m=149559827906211&w=2
> > [2] https://lore.kernel.org/lkml/20170606105707.23207-1-pbonzini@redhat.com/
> >
> > Paolo Bonzini (3):
> >   KVM: VMX: extract __pi_post_block
> >   KVM: VMX: avoid double list add with VT-d posted interrupts
> >   KVM: VMX: simplify and fix vmx_vcpu_pi_load
> >
> >  arch/x86/kvm/vmx.c | 177 ++++++++++++++++++++++-----------------------
> >  1 file changed, 87 insertions(+), 90 deletions(-)
> >
>
> The first patch appears like it can be safely ignored from a change point of
> view as it effectively just moves a code section into its own function.
> The other 2 are rather obscure in what is getting changed. What I mean is that
> its hard to tell what this code does anyway. So I would rely more on the
> successful testing. On the regression potential: if you have more insights on
> how a regression might show up, that should be rather mentioned. Very vaguely
> this changes around vCPU hotplug(?), so in that case regression potential would
> be issues in other cases of vCPU addition/removal...
>
> Acked-by: Stefan Bader <stefan.bader at canonical.com>
>

Thanks for the feedback. I cannot reproduce the bug myself. The user
that reported the bug
has tested and so far they don't have mentioned any regression.
FYI, after exchanging a couple of emails with William I sent a version
3 of the patches on Jan 18
with subject :
[SRU][Xenial][PATCH v3 0/4] Fix corruption on blocked_vcpu_on_cpu list

I don't know how you handle these cases in the kernel team. Both v2
and v3 are functionally
the same, in v3 the backport is more straightforward because I pull in
an extra commit.