linux-kernel - Re: [PATCH] of/irq: parse interrupts-extended during irq init heirarchy calculation

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: <b9c97f5e-ca61-ee91-0cb7-a426a921474c@sifive.com>
Date:   Mon, 11 Jul 2022 08:16:57 +0100
From:   Ben Dooks <ben.dooks@...ive.com>
To:     Rob Herring <robh+dt@...nel.org>
Cc:     devicetree@...r.kernel.org, Frank Rowand <frowand.list@...il.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Sudip Mukherjee <sudip.mukherjee@...ive.com>,
        Jude Onyenegecha <jude.onyenegecha@...ive.com>
Subject: Re: [PATCH] of/irq: parse interrupts-extended during irq init
 heirarchy calculation

On 08/07/2022 22:49, Rob Herring wrote:
> On Fri, Jul 8, 2022 at 10:52 AM Ben Dooks <ben.dooks@...ive.com> wrote:
>>
>> When the irq controler code works out the heirarchy for initialialisation
>> it only looks at interrupt-parent properties, but controllers such as the
>> RISC-V PLIC use a extended-interrupt property and therefore do not get
>> properly considered during initialisation.
>>
>> This means that if anything changes in the driver initialisation order
>> then the PLIC can get called before the CLINT nodes, and thus interrupts
>> do not get configured properly and the init continues without noticing
>> the error until drivers fail due to having no interrupts delivered.
>>
>> Add code to the of_irq_init that checks for the extended-interrupt
>> property and adds these parent nodes so that they can be considered
>> during the calculations of whether an irq controller node can be
>> initialised.
> 
> Isn't this already fixed by commit e91033621d56 ("of/irq: Use
> interrupts-extended to find parent")?

I didn't notice that commit, have been holding onto some of these
patches whilst sorting out permission to send.

The only thing my version does better is that it checks all of the
interrupt-extended contents so if there was a system with multiple
parents then we wouldn't consider them.

-- 
Ben