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: <aJDQHMbiNVToTbum@J2N7QTR9R3>
Date: Mon, 4 Aug 2025 16:22:04 +0100
From: Mark Rutland <mark.rutland@....com>
To: Wei-Lin Chang <r09922117@...e.ntu.edu.tw>
Cc: Anshuman Khandual <anshuman.khandual@....com>,
	linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org, Marc Zyngier <maz@...nel.org>,
	Oliver Upton <oliver.upton@...ux.dev>,
	Joey Gouly <joey.gouly@....com>,
	Sebastian Ene <sebastianene@...gle.com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Zenghui Yu <yuzenghui@...wei.com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>
Subject: Re: [PATCH] KVM: arm64: ptdump: Fix exec attribute printing

On Mon, Aug 04, 2025 at 08:41:35PM +0800, Wei-Lin Chang wrote:
> Hi Anshuman,
> 
> On Sun, Aug 03, 2025 at 07:33:04PM +0530, Anshuman Khandual wrote:
> > 
> > 
> > On 02/08/25 4:10 PM, Wei-Lin Chang wrote:
> > > Currently the guest stage-2 page table dump has the executable attribute
> > > printed in reverse, showing "X" for a non-executable region and showing
> > > " " for an executable one. This is caused by misjudgement of which
> > > string gets printed for the executable and non-executable case. Fix it
> > > by swapping the two strings.
> > > 
> > > Signed-off-by: Wei-Lin Chang <r09922117@...e.ntu.edu.tw>
> > > ---
> > >  arch/arm64/kvm/ptdump.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/arch/arm64/kvm/ptdump.c b/arch/arm64/kvm/ptdump.c
> > > index 098416d7e5c25..99fc13f1c11fb 100644
> > > --- a/arch/arm64/kvm/ptdump.c
> > > +++ b/arch/arm64/kvm/ptdump.c
> > > @@ -44,8 +44,8 @@ static const struct ptdump_prot_bits stage2_pte_bits[] = {
> > >  	}, {
> > >  		.mask	= KVM_PTE_LEAF_ATTR_HI_S2_XN | PTE_VALID,
> > >  		.val	= PTE_VALID,
> > > -		.set	= " ",
> > > -		.clear	= "X",
> > > +		.set	= "X",
> > > +		.clear	= " ",
> > >  	}, {

I think the big problem here is that we've included the 'PTE_VALID' bit
in the mask. We don't bother with that for the Stage-1 ptdump code, e.g.

        {
                .mask   = PTE_PXN,
                .val    = PTE_PXN,
                .set    = "NX",
                .clear  = "x ",
        },

.... so do we actually need to take the PTE_VALID bit into account here? Do
invalid Stage-2 entries have anything we don't want to report?

... or can we change the Stage-2 ptdump code to have:

	{
		.mask	= KVM_PTE_LEAF_ATTR_HI_S2_XN,
		.val	= KVM_PTE_LEAF_ATTR_HI_S2_XN,
		.set	= "NX",
		.clear	= "x ",
	},

... and match the Stage-1 code?

Otherwise, maybe we can add a separate valid-only filter.

> > >  		.mask	= KVM_PTE_LEAF_ATTR_LO_S2_AF | PTE_VALID,
> > >  		.val	= KVM_PTE_LEAF_ATTR_LO_S2_AF | PTE_VALID,
> > 
> > Is not KVM_PTE_LEAF_ATTR_HI_S2_XN already in the reverse semantics aka
> > XN (Execute Never). Hence when KVM_PTE_LEAF_ATTR_HI_S2_XN macro is set
> > that means the entry is not executable and vice versa.
> 
> Yes you are correct. However in dump_prot() we have:
> 
> if ((st->current_prot & bits->mask) == bits->val)
> 	s = bits->set;
> else
> 	s = bits->clear;
> 
> Analysis:
> 
> 1. region is executable:
>     - st->current_prot == PTE_VALID (ignore other bits)
>     - st->current_prot & bits->mask gets PTE_VALID
>     - if condition is true (.val is PTE_VALID)
>     - prints bits->set
> 
> 2. region is not executable:
>     - st->current_prot == KVM_PTE_LEAF_ATTR_HI_S2_XN | PTE_VALID
>     - st->current_prot & bits->mask gets (KVM_PTE_LEAF_ATTR_HI_S2_XN | PTE_VALID)
>     - if condition is false
>     - prints bits->clear

As above, this is because the PTE_VALID bit has been placed into the
mask, and that's not how the ptdump code was intended to be used.

> Therefore we want .set = "X", and .clear = " ".

That'll work around the problem, but I'm not sure that's the right fix.
If nothing else, it's *very* confusing.

Mark.

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ