linux-kernel - Re: [PATCH 27/28] KVM: x86/mmu: Do remote TLB flush before dropping RCU in TDP MMU resched

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <YaZG/NopJ7YaVUjD@google.com>
Date:   Tue, 30 Nov 2021 15:45:00 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Paolo Bonzini <pbonzini@...hat.com>
Cc:     Vitaly Kuznetsov <vkuznets@...hat.com>,
        Wanpeng Li <wanpengli@...cent.com>,
        Jim Mattson <jmattson@...gle.com>,
        Joerg Roedel <joro@...tes.org>, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Hou Wenlong <houwenlong93@...ux.alibaba.com>,
        Ben Gardon <bgardon@...gle.com>
Subject: Re: [PATCH 27/28] KVM: x86/mmu: Do remote TLB flush before dropping
 RCU in TDP MMU resched

On Tue, Nov 30, 2021, Paolo Bonzini wrote:
> On 11/20/21 05:50, Sean Christopherson wrote:
> >   	if (need_resched() || rwlock_needbreak(&kvm->mmu_lock)) {
> > -		rcu_read_unlock();
> > -
> >   		if (flush)
> >   			kvm_flush_remote_tlbs(kvm);
> > +		rcu_read_unlock();
> > +
> 
> Couldn't this sleep in kvm_make_all_cpus_request, whilst in an RCU read-side
> critical section?

No.  And if kvm_make_all_cpus_request() can sleep, the TDP MMU is completely hosed
as tdp_mmu_zap_spte_atomic() and handle_removed_tdp_mmu_page() currently call
kvm_flush_remote_tlbs_with_range() while under RCU protection.

kvm_make_all_cpus_request_except() disables preemption via get_cpu(), and
smp_call_function() doubles down on disabling preemption as the inner helpers
require preemption to be disabled, so anything below them should complain if
there's a might_sleep().  hv_remote_flush_tlb_with_range() takes a spinlock, so
nothing in there should be sleeping either.