linux-kernel - Re: [PATCH] KVM: arm64: nv: Fix s_cpu_if->vgic_lr[] indexing in vgic_v3_put

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: <aFD0wHqZ-8CRWIW-@linux.dev>
Date: Mon, 16 Jun 2025 21:53:20 -0700
From: Oliver Upton <oliver.upton@...ux.dev>
To: Marc Zyngier <maz@...nel.org>
Cc: Wei-Lin Chang <r09922117@...e.ntu.edu.tw>,
	linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org, Joey Gouly <joey.gouly@....com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Zenghui Yu <yuzenghui@...wei.com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>,
	Jintack Lim <jintack@...columbia.edu>,
	Christoffer Dall <christoffer.dall@....com>
Subject: Re: [PATCH] KVM: arm64: nv: Fix s_cpu_if->vgic_lr[] indexing in
 vgic_v3_put_nested()

On Mon, Jun 16, 2025 at 11:54:57AM +0100, Marc Zyngier wrote:
> On Sat, 14 Jun 2025 15:57:21 +0100,
> Wei-Lin Chang <r09922117@...e.ntu.edu.tw> wrote:
> > 
> > s_cpu_if->vgic_lr[] is filled continuously from index 0 to
> > s_cpu_if->used_lrs - 1, but vgic_v3_put_nested() is indexing it using
> > the positions of the set bits in shadow_if->lr_map. So correct it.
> > 
> > Signed-off-by: Wei-Lin Chang <r09922117@...e.ntu.edu.tw>
> > ---
> >  arch/arm64/kvm/vgic/vgic-v3-nested.c | 7 ++++---
> >  1 file changed, 4 insertions(+), 3 deletions(-)
> > 
> > diff --git a/arch/arm64/kvm/vgic/vgic-v3-nested.c b/arch/arm64/kvm/vgic/vgic-v3-nested.c
> > index 4f6954c30674..29741e3f077b 100644
> > --- a/arch/arm64/kvm/vgic/vgic-v3-nested.c
> > +++ b/arch/arm64/kvm/vgic/vgic-v3-nested.c
> > @@ -343,7 +343,7 @@ void vgic_v3_put_nested(struct kvm_vcpu *vcpu)
> >  	struct shadow_if *shadow_if = get_shadow_if();
> >  	struct vgic_v3_cpu_if *s_cpu_if = &shadow_if->cpuif;
> >  	u64 val;
> > -	int i;
> > +	int i, index = 0;
> >  
> >  	__vgic_v3_save_vmcr_aprs(s_cpu_if);
> >  	__vgic_v3_deactivate_traps(s_cpu_if);
> > @@ -368,10 +368,11 @@ void vgic_v3_put_nested(struct kvm_vcpu *vcpu)
> >  		val = __vcpu_sys_reg(vcpu, ICH_LRN(i));
> >  
> >  		val &= ~ICH_LR_STATE;
> > -		val |= s_cpu_if->vgic_lr[i] & ICH_LR_STATE;
> > +		val |= s_cpu_if->vgic_lr[index] & ICH_LR_STATE;
> >  
> >  		__vcpu_sys_reg(vcpu, ICH_LRN(i)) = val;
> > -		s_cpu_if->vgic_lr[i] = 0;
> > +		s_cpu_if->vgic_lr[index] = 0;
> > +		index++;
> >  	}
> >  
> >  	shadow_if->lr_map = 0;
> 
> Nice catch, thanks a lot for tracking it down.
> 
> However, I think we should get rid of this double-indexing altogether,
> or at least make it less error-prone. This thing is extremely fragile,
> and it isn't the first time we are getting bitten with it.
> 
> Looking at the code, it becomes pretty obvious that the shadow index
> is always the number of bits set in lr_map, and that we could
> completely drop the 'index' thing if we simply counted these bits
> (which isn't that expensive).
> 
> I came up with the (admittedly much bigger) following fix.
> 
> Thoughts?
> 
> 	M.
> 
> From 2484950b8fc3b36cca32bf5e86ffe7975a43e0e7 Mon Sep 17 00:00:00 2001
> From: Marc Zyngier <maz@...nel.org>
> Date: Sun, 15 Jun 2025 16:11:38 +0100
> Subject: [PATCH] KVM: arm64: nv: Fix tracking of shadow list registers
> 
> Wei-Lin reports that the tracking of shadow list registers is
> majorly broken when resync'ing the L2 state after a run, as
> we confuse the guest's LR index with the host's, potentially
> losing the interrupt state.
> 
> While this could be fixed by adding yet another side index to
> track it (Wei-Lin's fix), it may be better to refactor this
> code to avoid having a side index altogether, limiting the
> risk to introduce this class of bugs.
> 
> A key observation is that the shadow index is always the number
> of bits in the lr_map bitmap. With that, the parallel indexing
> scheme can be completely dropped.
> 
> While doing this, introduce a couple of helpers that abstract
> the index conversion and some of the LR repainting, making the
> whole exercise much simpler.
> 
> Reported-by: Wei-Lin Chang <r09922117@...e.ntu.edu.tw>
> Signed-off-by: Marc Zyngier <maz@...nel.org>
> Link: https://lore.kernel.org/r/20250614145721.2504524-1-r09922117@csie.ntu.edu.tw

Besides Wei-Lin's comments, LGTM.

Reviewed-by: Oliver Upton <oliver.upton@...ux.dev>

Thanks,
Oliver