linux-kernel - Re: fwnode_for_each_child_node() and OF backend discrepancy

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: <2f2d7685e0e43194270a310034004970@walle.cc>
Date:   Tue, 28 Jun 2022 15:23:50 +0200
From:   Michael Walle <michael@...le.cc>
To:     Andy Shevchenko <andy.shevchenko@...il.com>
Cc:     Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        ACPI Devel Maling List <linux-acpi@...r.kernel.org>,
        devicetree <devicetree@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: fwnode_for_each_child_node() and OF backend discrepancy

>> I was trying to fix the lan966x driver [1] which doesn't work if there
>> are disabled nodes in between.
> 
> Can you elaborate what's wrong now in the behaviour of the driver? In
> the code it uses twice the _available variant.

Imagine the following device tree snippet:
  port0 {
    reg = <0>;
    status = "okay";
  }
  port1 {
    reg = <1>;
    status = "disabled";
  }
  port@2 {
    reg = <2>;
    status = "okay";
  }

The driver will set num_phys_ports to 2. When port@2 is probed, it
will have the (correct!) physical port number 2. That will then
trigger various EINVAL checks with "port_num >= num_phys_ports" or
WARN()s.

So the easiest fix would be to actual count all the child nodes
(regardless if they are available or not), assuming there are as
many nodes as physical ports.

But num_phys_ports being a property of the hardware I don't
think it's good to deduce it by counting the child nodes anyway,
but it should rather be a (hardcoded) property of the driver.

-michael

[1] 
https://elixir.bootlin.com/linux/v5.19-rc4/source/drivers/net/ethernet/microchip/lan966x/lan966x_main.c