linux-kernel - Re: [PATCH 0/4] of: reserved-memory: Various improvements

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]

Date:   Mon, 6 Apr 2020 12:04:11 +0200
From:   Thierry Reding <thierry.reding@...il.com>
To:     Rob Herring <robh+dt@...nel.org>
Cc:     Frank Rowand <frowand.list@...il.com>, devicetree@...r.kernel.org,
        linux-tegra <linux-tegra@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 0/4] of: reserved-memory: Various improvements

On Sat, Apr 04, 2020 at 07:24:25PM -0600, Rob Herring wrote:
> On Fri, Apr 3, 2020 at 12:56 PM Thierry Reding <thierry.reding@...il.com> wrote:
> >
> > From: Thierry Reding <treding@...dia.com>
> >
> > Hi Rob, all,
> >
> > this is a set of patches that I've been working on to allow me to use
> > reserved memory regions more flexibly. One of the use-cases that I have
> > is an external memory controller driver that gets passed one or two
> > tables from firmware containing a set of EMC frequencies and the
> > corresponding register values to program for these frequencies.
> >
> > One of these tables is the "nominal" table and an optional second table
> > is "derated" and is used when the DRAM chips are overheating. I want to
> > be able to pass these tables as separate memory-region entries.
> >
> > So what this small patchset does is make the reserved-memory code adapt
> > to this situation better. On one hand, while the DT bindings currently
> > support multiple regions per device tree node, it's slightly unintuitive
> > to specify them. The first patch adds a memory-region-names property
> > that allows the DT to specify a "consumer" name for these regions much
> > like we do for things like clocks, resets or the reg property. At the
> > same time, a new alias for memory-region, named memory-regions, is
> > introduced to make this more consistent with other bindings.
> 
> It's just not worth supporting both flavors (forever). I don't want to
> repeat gpio vs. gpios. Let's just stick with 'memory-region' and allow
> that to be more than one entry.

Alright, I'll drop the corresponding changes from the bindings and the
OF core then.

> I'm not a fan of *-names, but fine.

I suppose I could work without them, but I like the descriptiveness that
they add to the device tree. There are also cases where they can be very
essential. For example, what if a device can take two separate memory
regions. One case that we have on Tegra is the display controller that
can have a framebuffer and a color-conversion lookup table assigned to
it on boot. I think in this particular case both are actually always
there, but what if either of them was optional. Without -names it would
be impossible to describe the context if only one is provided in device
tree.

Thierry

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)