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  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]
Date:   Sat, 16 Jan 2021 22:25:23 +0100
From:   Marek Vasut <marex@...x.de>
To:     Lukas Wunner <lukas@...ner.de>, Arnd Bergmann <arnd@...nel.org>
Cc:     Networking <netdev@...r.kernel.org>, Andrew Lunn <andrew@...n.ch>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Jakub Kicinski <kuba@...nel.org>
Subject: Re: [PATCH net-next V2] net: ks8851: Fix mixed module/builtin build

On 1/16/21 9:39 PM, Lukas Wunner wrote:
> On Sat, Jan 16, 2021 at 08:26:22PM +0100, Arnd Bergmann wrote:
>> On Sat, Jan 16, 2021 at 6:56 PM Marek Vasut <marex@...x.de> wrote:
>>> On 1/16/21 6:04 PM, Arnd Bergmann wrote:
>>>> On Sat, Jan 16, 2021 at 5:48 PM Marek Vasut <marex@...x.de> wrote:
>>>
>>>> I don't really like this version, as it does not actually solve the problem of
>>>> linking the same object file into both vmlinux and a loadable module, which
>>>> can have all kinds of side-effects besides that link failure you saw.
>>>>
>>>> If you want to avoid exporting all those symbols, a simpler hack would
>>>> be to '#include "ks8851_common.c" from each of the two files, which
>>>> then always duplicates the contents (even when both are built-in), but
>>>> at least builds the file the correct way.
>>>
>>> That's the same as V1, isn't it ?
>>
>> Ah, I had not actually looked at the original submission, but yes, that
>> was slightly better than v2, provided you make all symbols static to
>> avoid the new link error.
>>
>> I still think that having three modules and exporting the symbols from
>> the common part as Heiner Kallweit suggested would be the best
>> way to do it.
> 
> FWIW I'd prefer V1 (the #include approach) as it allows going back to
> using static inlines for register access.  That's what we had before
> 7a552c850c45.
> 
> It seems unlikely that a system uses both, the parallel *and* the SPI
> variant of the ks8851.  So the additional memory necessary because of
> code duplication wouldn't matter in practice.

I have a board with both options populated on my desk, sorry.

Powered by blists - more mailing lists