Re: [PATCH] KVM: avoid warning on s390 in mark_page_dirty

[Christian Borntraeger]

Am 18.01.22 um 09:44 schrieb Paolo Bonzini:
> On 1/18/22 09:37, Christian Borntraeger wrote:
> > Am 13.01.22 um 13:29 schrieb Christian Borntraeger:
> > > Avoid warnings on s390 like
> > > [ 1801.980931] CPU: 12 PID: 117600 Comm: kworker/12:0 Tainted: G            E 5.17.0-20220113.rc0.git0.32ce2abb03cf.300.fc35.s390x+next #1
> > > [ 1801.980938] Workqueue: events irqfd_inject [kvm]
> > > [...]
> > > [ 1801.981057] Call Trace:
> > > [ 1801.981060]  [<000003ff805f0f5c>] mark_page_dirty_in_slot+0xa4/0xb0 [kvm]
> > > [ 1801.981083]  [<000003ff8060e9fe>] adapter_indicators_set+0xde/0x268 [kvm]
> > > [ 1801.981104]  [<000003ff80613c24>] set_adapter_int+0x64/0xd8 [kvm]
> > > [ 1801.981124]  [<000003ff805fb9aa>] kvm_set_irq+0xc2/0x130 [kvm]
> > > [ 1801.981144]  [<000003ff805f8d86>] irqfd_inject+0x76/0xa0 [kvm]
> > > [ 1801.981164]  [<0000000175e56906>] process_one_work+0x1fe/0x470
> > > [ 1801.981173]  [<0000000175e570a4>] worker_thread+0x64/0x498
> > > [ 1801.981176]  [<0000000175e5ef2c>] kthread+0x10c/0x110
> > > [ 1801.981180]  [<0000000175de73c8>] __ret_from_fork+0x40/0x58
> > > [ 1801.981185]  [<000000017698440a>] ret_from_fork+0xa/0x40
> > >
> > > when writing to a guest from an irqfd worker as long as we do not have
> > > the dirty ring.
> > >
> > > Signed-off-by: Christian Borntraeger <borntraeger@xxxxxxxxxxxxx>
> > > ---
> > >   virt/kvm/kvm_main.c | 2 ++
> > >   1 file changed, 2 insertions(+)
> > >
> > > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > > index 504158f0e131..1a682d3e106d 100644
> > > --- a/virt/kvm/kvm_main.c
> > > +++ b/virt/kvm/kvm_main.c
> > > @@ -3163,8 +3163,10 @@ void mark_page_dirty_in_slot(struct kvm *kvm,
> > >   {
> > >       struct kvm_vcpu *vcpu = kvm_get_running_vcpu();
> > > +#ifdef CONFIG_HAVE_KVM_DIRTY_RING
> > >       if (WARN_ON_ONCE(!vcpu) || WARN_ON_ONCE(vcpu->kvm != kvm))
> > >           return;
> > > +#endif
> > >       if (memslot && kvm_slot_dirty_track_enabled(memslot)) {
> > >           unsigned long rel_gfn = gfn - memslot->base_gfn;
> >
> > Paolo, are you going to pick this for next for the time being?
>
> Yep, done now.
>
> Paolo

Thanks. I just realized that Davids patch meanwhile landed in Linus tree. So better
take this via master and not next.
Maybe also add
Fixes: 2efd61a608b0 ("KVM: Warn if mark_page_dirty() is called without an active vCPU")
in case the patch is picked for stable