linux-kernel - Re: [Cocci] [PATCH v5 0/3] Add error message to platform_get

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: <5d41ab2c.1c69fb81.6129.661f@mx.google.com>
Date:   Wed, 31 Jul 2019 07:52:27 -0700
From:   Stephen Boyd <swboyd@...omium.org>
To:     Wolfram Sang <wsa@...-dreams.de>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Rob Herring <robh@...nel.org>,
        Michal Marek <michal.lkml@...kovi.net>,
        Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
        "Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
        Nicolas Palix <nicolas.palix@...g.fr>,
        linux-kernel@...r.kernel.org,
        Javier Martinez Canillas <javierm@...hat.com>,
        Andrzej Hajda <a.hajda@...sung.com>,
        Andy Shevchenko <andy.shevchenko@...il.com>,
        Mark Brown <broonie@...nel.org>,
        Russell King - ARM Linux <linux@...linux.org.uk>,
        cocci@...teme.lip6.fr, Marek Szyprowski <m.szyprowski@...sung.com>
Subject: Re: [Cocci] [PATCH v5 0/3] Add error message to platform_get_irq*()

Quoting Wolfram Sang (2019-07-31 07:26:46)
> Hi Stephen,
> 
> > There were some comments about adding an 'optional' platform_get_irq()
> > API in v4. This series doesn't include that, but I can add such an API
> > if it's required. I started to look into how it might work and got hung
> > up on what an optional IRQ means. I suppose it means that in DT there
> > isn't an 'interrupts' property in the device node, but in ACPI based
> > firmware I'm not sure what that would correspond to. Furthermore, the
> > return value is hard to comprehend. Do we return an error when an
> > optional irq can't be found? It doesn't seem safe to return 0 because
> > sometimes 0 is a valid IRQ. Do other errors in parsing the IRQ
> > constitute a failure when the IRQ is optional?
> 
> Some time ago, I tried a series like yours and got stuck at this very
> point. I found drivers where using an interrupt was optional and
> platform_get_irq() returning a failure wasn't fatal. The drivers used
> PIO then or dropped some additional functionality. Some of them were
> very old.
> 
> I didn't like the idea that platform_get_irq() will spit out errors for
> those drivers, yet I couldn't create a suitable cocci-script to convert
> drivers to use the *_optional callback where possible. So, I neither
> created the optional callback.
> 
> I still have doubts of unneeded error messages popping up. Has this been
> discussed already? (Sorry, I missed the first iterations of this series)

Not heavily discussed, but it was mentioned in an earlier round. If
these drivers pop up, I think we can have another function like
platform_get_irq_probe() or platform_get_irq_nowarn() that doesn't print
an error message. Then we can convert the drivers that are poking around
for interrupts to use this new function instead. It isn't the same as a
platform_get_optional_irq() API because it returns an error when the irq
isn't there or we fail to parse something, but at least the error
message is gone.