linux-kernel - Re: [PATCH v2] ARM: dt: check MPIDR on MP devices built without SMP

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] [day] [month] [year] [list]

Message-ID: <061822b6ad80094a52d27f27f3e37594adb313c2.camel@suse.de>
Date:   Mon, 18 Nov 2019 18:45:55 +0100
From:   Nicolas Saenz Julienne <nsaenzjulienne@...e.de>
To:     Russell King - ARM Linux admin <linux@...linux.org.uk>
Cc:     linux-kernel@...r.kernel.org, f.fainelli@...il.com,
        "kernelci.org bot" <bot@...nelci.org>, wahrenst@....net,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v2] ARM: dt: check MPIDR on MP devices built without SMP

Hi Russell, thanks for the review.

On Mon, 2019-11-18 at 12:55 +0000, Russell King - ARM Linux admin wrote:
> On Mon, Nov 18, 2019 at 12:49:04PM +0100, Nicolas Saenz Julienne wrote:
> > On Fri, 2019-10-04 at 17:52 +0200, Nicolas Saenz Julienne wrote:
> > > On SMP builds, in order to properly link CPU devices with their
> > > respective DT nodes we start by matching the boot CPU. This is achieved
> > > by comparing the 'reg' property on each of the CPU DT nodes with the
> > > MPIDR. The association is necessary as to validate the whole CPU logical
> > > map, which ultimately links CPU devices and their DT nodes.
> 
> No, that is not the primary purpose of the CPU logical map.  The CPU 
> logical map is there to map the CPU logical number to a hardware number,
> necessary for programming hardware.

Noted.

> > > On setups built without SMP, no MPIDR read is performed. The only thing
> > > expected is for the 'reg' property in the CPU DT node to contain the
> > > value 0x0.
> > > 
> > > This causes problems on MP setups built without SMP. As their boot CPU
> > > DT node contains the relevant MPIDR as opposed to 0x0. No match is then
> > > possible. This causes troubles further down the line as drivers are
> > > unable to get the CPU's DT node.
> 
> So the DT is incorrect for the platform - it is not describing the
> hardware.  Why can't the DT be fixed?  Clearly, it would have never
> worked with the mainline kernel today.

Sorry but I don't see any incorrect DT here. From the ARM CPU bindings I gather
that (at least since ARMv7) every CPU node should contain its corresponding
MPIDR. It transpires that ARM's DT cpu map init code should take that into
account regardles of whether the kernel supports SMP, isn't it?

Regards,
Nicolas


Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)