linux-kernel - Re: [PATCH 03/14] KVM: arm64: Continue stage-2 map when re-creating mappings

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 for Android: free password hash cracker in your pocket

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

Message-ID: <YPa6BGuUFjw8do+o@google.com>
Date:   Tue, 20 Jul 2021 12:56:52 +0100
From:   Quentin Perret <qperret@...gle.com>
To:     Marc Zyngier <maz@...nel.org>
Cc:     james.morse@....com, alexandru.elisei@....com,
        suzuki.poulose@....com, catalin.marinas@....com, will@...nel.org,
        linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.cs.columbia.edu,
        linux-kernel@...r.kernel.org, ardb@...nel.org, qwandor@...gle.com,
        tabba@...gle.com, dbrazdil@...gle.com, kernel-team@...roid.com,
        Yanan Wang <wangyanan55@...wei.com>
Subject: Re: [PATCH 03/14] KVM: arm64: Continue stage-2 map when re-creating
 mappings

On Tuesday 20 Jul 2021 at 09:26:10 (+0100), Marc Zyngier wrote:
> Right, but this is on a different path, right? Guests can never fault
> multiple mappings at once, and it takes you a host hypercall to
> perform this "multiple leaves at once".

Right.

> Is there any way we can restrict this to the hypercall? Or even
> better, keep the hypercall as a "one page at a time" thing? I can't
> imagine it being performance critical (it is a once-off, and only used
> over a rather small region of memory). Then, the called doesn't have
> to worry about the page already being mapped or not. This would also
> match the behaviour of what I do on the MMIO side.
> 
> Or do you anticipate much gain from this being able to use block
> mappings?

Not really no, especially given that mappings of shared pages will be
forced to page granularity thanks to the other patch we discussed in
this series. I was just hoping to reduce the overhead a bit by reducing
the number of hypercalls. But as you said, this probably doesn't matter
all that much, so happy to rework that. I'll look into making the hcall
use one page at a time, and hopefully that'll simplify a few other
things in the check_host_share_hyp() path near the end of this series.

Thanks,
Quentin