linux-kernel - Re: [PATCH v1] KVM: x86: PMU Whitelist

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]

Date:   Tue, 2 Jul 2019 10:46:16 -0700
From:   Eric Hankland <ehankland@...gle.com>
To:     Wei Wang <wei.w.wang@...el.com>
Cc:     Paolo Bonzini <pbonzini@...hat.com>, rkrcmar@...hat.com,
        linux-kernel@...r.kernel.org, kvm@...r.kernel.org
Subject: Re: [PATCH v1] KVM: x86: PMU Whitelist

> Actually I have another thing to discuss:
> probably we could consider to make this filter list white/black configurable
> from userspace. For example, userspace option: filter-list=white/black

Works for me. I'll include this in the next version.

> Probably we don't need this field to be passed from userspace?
>
> We could directly use AMD64_RAW_EVENTMASK_NB, which includes bit[35:32].
> Since those bits are reserved on Intel CPUs, have them as mask should be
> fine.
>
> Alternatively, we could add this event_mask field to struct kvm_pmu, and
> initalize
> it in the vendor specific intel_pmu_init or amd_pmu_init.
>
> Both options above look good to me.

Sounds good. I'll go with the first suggestion for now since it's
simpler. If other reviewers prefer the second I can implement that.

> For the fixed counter, we could add a bitmap flag to kvm_arch,
> indicating which counter is whitelist-ed based on the
> "eventsel+umask" value passed from userspace. This flag is
> updated when updating the whitelist-ed events to kvm.
> For example, if userspace gives "00+01" (INST_RETIRED_ANY),
> then we enable fixed counter0 in the flag.
>
> When reprogram_fixed_counter, we check the flag and return
> if the related counter isn't whitelisted.

Sounds good to me.

If you don't have any more comments I'll send out the next version
with all the requested changes.

Eric