linux-kernel - Some questions about DT-schema

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

Date:   Fri, 21 Feb 2020 15:40:48 +0900
From:   Masahiro Yamada <masahiroy@...nel.org>
To:     Rob Herring <robh+dt@...nel.org>,
        Frank Rowand <frowand.list@...il.com>,
        DTML <devicetree@...r.kernel.org>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Some questions about DT-schema

[Q1] Order of reset-names, clock-names, etc.

I have nodes that describe multiple reset lines:

It looks like this:

    reset-names = "host", "bridge";
    resets = <&rst 0>, <&rst 1>;

I looks into some existing schema files
to figure out how to describe this.

We typically do this.

reset-names:
  - items:
    - const: host
    - const: bridge

Since 'items' checks each item against the corresponding index,
it is order-sensitive.

So, the following DT passes the schema.

    reset-names = "host", "bridge";
    resets = <&rst 0>, <&rst 1>;

However, the following DT fails in the schema checking.

    reset-names = "bridge", "host";
    resets = <&rst 1>, <&rst 0>;

The latter is as correct as the former,
and should work equivalently.

If we are always required to write the "host", "bridge", in this order,
there is no point of 'reset-names'.
So, this is a restriction we should not impose.

So, if we want to accept both cases,
we need to write like this:

reset-names:
  - oneOf:
    - items:
      - const: host
      - const: bridge
    - items:
      - const: bridge
      - const: host

If we have 3 reset singles, we end up with listing 6 patterns.
Is there a good way to describe this?

Or, the policy is,
there is only one way to do one thing ?

[Q2] Tupling reg, range, etc.

In the context of dt-schema,
'reg' is essentially, array of array,
and it is important to how you tuple values.

    reg = <1 2>, <3 4>;

    reg = <1>, <2>, <3>, <4>;

    reg = <1 2 3 4>;

All of the three are compiled into the equivalent DTB,
but in the context of schema checking,
the number of items is, 2, 4, 1, respectively.

So, we need to care about tuple values correctly
based on #address-cells and #size-cells.

In some DT, I previously wrote ranges like this

   ranges = <1 0x00000000 0x42000000 0x02000000,
             5 0x00000000 0x46000000 0x01000000>;

But, now probably more correct way is:

   ranges = <1 0x00000000 0x42000000 0x02000000>,
            <5 0x00000000 0x46000000 0x01000000>;

This is a new restriction, and it also means
we cannot perform schema checking against
dis-assembled DT  (dtb->dts).

Is this correct?

-- 
Best Regards
Masahiro Yamada