linux-kernel - Re: [PATCH] arm64/cpuinfo: Drop boot_cpu

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: <224296d1-086a-5516-95a8-8f4ad5c533d9@arm.com>
Date:   Mon, 4 May 2020 20:23:08 +0530
From:   Anshuman Khandual <anshuman.khandual@....com>
To:     Mark Rutland <mark.rutland@....com>
Cc:     linux-arm-kernel@...ts.infradead.org,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>, Mark Brown <broonie@...nel.org>,
        Suzuki Poulose <suzuki.poulose@....com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] arm64/cpuinfo: Drop boot_cpu_data

On 05/04/2020 06:13 PM, Mark Rutland wrote:
> On Mon, May 04, 2020 at 06:00:00PM +0530, Anshuman Khandual wrote:
>> A global boot_cpu_data is not really required. Lets drop this.
> 
> I don't think it's true that this isn't required today.
> 
> One reason that we have both boot_cpu_data and a cpu_data variable for
> CPU0 is that CPU0 itself can be hotplugged out then back in, and this
> allows us to detect if CPU0's features have changed (e.g. due to FW
> failing to configure it appropriately, or real physical hotplug
> occurring).

Understood. After hotplug, CPU0 will come back via secondary_start_kernel()
where it's current register values will be checked against earlier captured
values i.e boot_cpu_data.

But wondering why should CPU0 be treated like any other secondary CPU. IOW
in case the fresh boot CPU register values dont match with boot_cpu_data,
should not the online process just be declined ? AFAICS, current approach
will let the kernel run with taint in case of a mismatch.

> 
> So NAK to the patch as it stands. If we're certain we capture all of
> those details even without boot_cpu_data, then we should make other
> changes to make that clear (e.g. removing it as an argument to
> update_cpu_features()).

There might not be another way, unless we can override CPU0's cpu_data
variable when the boot CPU comes back in after vetting against existing
values. Is there any particular reason to store the very first boot CPU0
info for ever ?

Passing on CPU0's cpu_data variable in update_cpu_features() for secondary
CPUs during boot still make sense. It helps in finalizing register values.
Re-entering CPU0's test against boot_cpu_data seems different.