linux-kernel - Re: [PATCH v2 1/2] dt-bindings: clock: mediatek: replace unusable clock

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: <20230526-acorn-unwound-34816283b187@wendy>
Date:   Fri, 26 May 2023 12:39:37 +0100
From:   Conor Dooley <conor.dooley@...rochip.com>
To:     Alexandre Mergnat <amergnat@...libre.com>
CC:     Conor Dooley <conor@...nel.org>,
        Michael Turquette <mturquette@...libre.com>,
        Stephen Boyd <sboyd@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Conor Dooley <conor+dt@...nel.org>,
        Matthias Brugger <matthias.bgg@...il.com>,
        AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@...labora.com>,
        Chen-Yu Tsai <wenst@...omium.org>,
        Markus Schneider-Pargmann <msp@...libre.com>,
        <linux-clk@...r.kernel.org>, <devicetree@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <linux-mediatek@...ts.infradead.org>
Subject: Re: [PATCH v2 1/2] dt-bindings: clock: mediatek: replace unusable
 clock

On Fri, May 26, 2023 at 10:54:04AM +0200, Alexandre Mergnat wrote:
> On 25/05/2023 19:51, Conor Dooley wrote:
> > On Thu, May 25, 2023 at 04:50:27PM +0200, Alexandre Mergnat wrote:
> > > The “mcu_pm_bclk_ck_cg” clock is used by co-processors and should not be
> > > added to the kernel driver, otherwise the CPU just halt and the board is
> > > rebooted by the wathdog.
> > > 
> > > Instead, add the "aes_top0_bclk_ck_cg" missing clock to prevent
> > > re-shuffling index and then preserve the ABI.
> > 
> > How does this preserve the ABI exactly? Please describe exactly what you
> > mean by that.
> 
> I mean that reduce the impact of the change compared to the v1 where I've
> changed the index of the following defines to be clean.

Oh, you can't do that at all as you probably discovered!

> > Also, what about any other users of these definitions, outside of Linux?
> 
> The clock driver and bindings are only a couple of kernel versions old, I'm
> pretty sure no one is using it.

Pretty sure, or sure?

> Also, if someone use CLK_IFR_MCU_PM_BK
> define, I'm wondering how his CPU is working since Mediatek told me that
> shouldn't be used, and after some try, I confirm.

Maybe that person is actually using the index to make sure that the
clock at that index is left untouched.

> I've a question: If something is wrong in the binding, you don't fix it to
> avoid ABI change ?

I don't quite get what you mean by "wrong". These header files just
define a set of arbitrary meanings, since the clock numbers are really
just something that developers came up with rather than being lifted
from a TRM. They don't prescribe behaviour for each of these clocks, or
that these clocks should actually be used - just a simple "this number
means this clock".
It sounds more like a driver or devicetree is _using_ the number
incorrectly, but that does not make the binding wrong :)

> TBH, I just try to clean the binding. I can fix the driver index issue
> (patch 2/2) without fixing the binding if you prefer. But IMHO, keep an
> unusable define isn't great...

I, at least, would prefer that.

Thanks,
Conor.

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)