Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220407195908.633003-1-pgonda@google.com> <CAFNjLiXC0AdOw5f8Ovu47D==ex7F0=WN_Ocirymz4xL=mWvC5A@mail.gmail.com>
 <CAMkAt6r-Mc_YN-gVHuCpTj4E1EmcvyYpP9jhtHo5HRHnoNJAdA@mail.gmail.com>
 <CAMkAt6r+OMPWCbV_svUyGWa0qMzjj2UEG29G6P7jb6uH6yko2w@mail.gmail.com>
 <62e9ece1-5d71-f803-3f65-2755160cf1d1@redhat.com> <CAMkAt6q6YLBfo2RceduSXTafckEehawhD4K4hUEuB4ZNqe2kKg@mail.gmail.com>
 <4c0edc90-36a1-4f4c-1923-4b20e7bdbb4c@redhat.com> <CAMkAt6oL5qi7z-eh4z7z8WBhpc=Ow6WtcJA5bDi6-aGMnz135A@mail.gmail.com>
 <CAMkAt6rmDrZfN5DbNOTsKFV57PwEnK2zxgBTCbEPeE206+5v5w@mail.gmail.com>
 <0d282be4-d612-374d-84ba-067994321bab@redhat.com> <CAMkAt6ragq4OmnX+n628Yd5pn51qFv4qV20upGR6tTvyYw3U5A@mail.gmail.com>
 <8a2c5f8c-503c-b4f0-75e7-039533c9852d@redhat.com> <CAMkAt6qAW5zFyTAqX_Az2DT2J3KROPo4u-Ak1sC0J+UTUeFfXA@mail.gmail.com>
 <4afce434-ab25-66d6-76f4-3a987f64e88e@redhat.com>
In-Reply-To: <4afce434-ab25-66d6-76f4-3a987f64e88e@redhat.com>
From:   Peter Gonda <pgonda@google.com>
Date:   Fri, 29 Apr 2022 11:12:36 -0600
Message-ID: <CAMkAt6o8u9=H_kjr_xyRO05J=RDFUZRiTc_Bw-FFDKEUaiyp2Q@mail.gmail.com>
Subject: Re: [PATCH v3] KVM: SEV: Mark nested locking of vcpu->lock
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     John Sperbeck <jsperbeck@google.com>,
        kvm list <kvm@vger.kernel.org>,
        David Rientjes <rientjes@google.com>,
        Sean Christopherson <seanjc@google.com>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Apr 29, 2022 at 9:59 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 4/29/22 17:51, Peter Gonda wrote:
> >> No, you don't need any of this.  You can rely on there being only one
> >> depmap, otherwise you wouldn't need the mock releases and acquires at
> >> all.  Also the unlocking order does not matter for deadlocks, only the
> >> locking order does.  You're overdoing it. :)
> >
> > Hmm I'm slightly confused here then. If I take your original suggestion of:
> >
> >          bool acquired = false;
> >          kvm_for_each_vcpu(...) {
> >                  if (acquired)
> >                          mutex_release(&vcpu->mutex.dep_map,
> > _THIS_IP_);  <-- Warning here
> >                  if (mutex_lock_killable_nested(&vcpu->mutex, role)
> >                          goto out_unlock;
> >                  acquired = true;
> >
> > """
> > [ 2810.088982] =====================================
> > [ 2810.093687] WARNING: bad unlock balance detected!
> > [ 2810.098388] 5.17.0-dbg-DEV #5 Tainted: G           O
> > [ 2810.103788] -------------------------------------
>
> Ah even if the contents of the dep_map are the same for all locks, it
> also uses the *pointer* to the dep_map to track (class, subclass) ->
> pointer and checks for a match.
>
> So yeah, prev_cpu is needed.  The unlock ordering OTOH is irrelevant so
> you don't need to visit the xarray backwards.

Sounds good. Instead of doing this prev_vcpu solution we could just
keep the 1st vcpu for source and target. I think this could work since
all the vcpu->mutex.dep_maps do not point to the same string.

Lock:
         bool acquired = false;
         kvm_for_each_vcpu(...) {
                 if (mutex_lock_killable_nested(&vcpu->mutex, role)
                     goto out_unlock;
                acquired = true;
                 if (acquired)
                      mutex_release(&vcpu->mutex, role)
         }

Unlock

         bool acquired = true;
         kvm_for_each_vcpu(...) {
               if (!acquired)
                     mutex_acquire(&vcpu->mutex, role)
               mutex_unlock(&vcpu->mutex);
               acquired = false;
         }

So when locking we release all but the first dep_maps. Then when
unlocking we acquire all but the first dep_maps. Thoughts?

>
> Paolo
>