linux-kernel - Re: drivers: Inline code in devm_platform_ioremap

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: <714a38fe-a733-7264-bb06-d94bd58a245a@metux.net>
Date:   Tue, 18 Jun 2019 11:35:34 +0200
From:   "Enrico Weigelt, metux IT consult" <lkml@...ux.net>
To:     Markus Elfring <Markus.Elfring@....de>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Bartosz Golaszewski <bgolaszewski@...libre.com>
Cc:     cocci@...teme.lip6.fr, kernel-janitors@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Gilles Muller <Gilles.Muller@...6.fr>,
        Himanshu Jha <himanshujha199640@...il.com>,
        Linus Walleij <linus.walleij@...aro.org>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Michal Marek <michal.lkml@...kovi.net>,
        Nicolas Palix <nicolas.palix@...g.fr>
Subject: Re: drivers: Inline code in devm_platform_ioremap_resource() from two
 functions

On 18.06.19 07:37, Markus Elfring wrote:
>>> Two function calls were combined in this function implementation.
>>> Inline corresponding code so that extra error checks can be avoided here.
>>
>> What exactly is the purpose of this ?
> 
> I suggest to take another look at the need and relevance of involved
> error checks in the discussed function combination.

Sorry, don't have the time for guessing and trying to reproduce your
thoughts. That's why we have patch descriptions / commit messages.
It would be a lot easier for all of us if you just desribe the exact
problem you'd like to solve and your approach to do so.

>> Looks like a notable code duplication ...
> 
> This can be.

I doubt that code duplication is appreciated, as this increases the
maintenance overhead. (actually, we're usually trying to reduce that,
eg. by using lots of generic helpers).

>> I thought we usually try to reduce this, instead of introducing new ones.
> 
> Would you like to check the software circumstances once more
> for the generation of a similar code structure by a C compiler
> (or optimiser)?

As said: unfortunately, I don't have the time to do that - you'd have to
tell us, what exactly you've got in mind.

If it's just about some error checks which happen to be redundant in a
particular case, you'll have to show that this case is a *really* hot
path (eg. irq, syscall, scheduling, etc) - but I don't see that here.

What's the exact scenario you're trying to optimize ? Any actual
measurements on how your patch improves that ?

Look, I understand that you'd like to squeeze out maximum performance,
but this has to be practically maintainable. I could list a lot of
things that I don't need in particular use cases and would like to
introduce build knobs for, but I have to understand that maintainers
have to be pretty reluctant towards those things.

--mtx

-- 
Enrico Weigelt, metux IT consult
Free software and Linux embedded engineering
info@...ux.net -- +49-151-27565287