linux-kernel - Re: Usefulness of CONFIG_MACH

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: <DBGIIR.8W0JWG83D6FE2@crapouillou.net>
Date:   Tue, 20 Sep 2022 14:33:13 +0100
From:   Paul Cercueil <paul@...pouillou.net>
To:     "H. Nikolaus Schaller" <hns@...delico.com>
Cc:     linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-mips <linux-mips@...r.kernel.org>,
        MIPS Creator CI20 Development 
        <mips-creator-ci20-dev@...glegroups.com>
Subject: Re: Usefulness of CONFIG_MACH_JZ47*



Le mar., sept. 20 2022 at 14:31:38 +0200, H. Nikolaus Schaller 
<hns@...delico.com> a écrit :
> Hi Paul,
> 
>>  Am 20.09.2022 um 11:09 schrieb Paul Cercueil <paul@...pouillou.net>:
>> 
>>  Hi Nikolaus,
>> 
>>  Le mar., sept. 20 2022 at 08:31:30 +0200, H. Nikolaus Schaller 
>> <hns@...delico.com> a écrit :
>>>  Hi Paul,
>>>  it seems as if there aren't many places left over where the 
>>> MACH_JZ47* configs are still in use:
>>>  drivers/char/hw_ramdom/Kconfig
>>>  drivers/clk/ingenic/Kconfig
>>>  drivers/gpu/drm/ingenic/Kconfig
>>>  drivers/pinctrl/pinctrl-ingenic.c
>>>  Is it possible to get rid of them and just have 
>>> CONFIG_MACH_INGENIC_GENERIC?
>>>  This might simplify my defconfig for multiple machines.
>> 
>>  CONFIG_MIPS_GENERIC_KERNEL=y
> 
> This breaks compilation for me, e.g.
> 
> arch/mips/mm/cache.c:203:6: error: 'cpu_has_tx39_cache' undeclared 
> (first use in this function)

v6.0-rc does not have 'cpu_has_tx39_cache' anywhere in that file, or in 
arch/mips/ for that matter. It was removed in v5.18.

And a v5.17 kernel compiles fine here with these options enabled. So 
it's a problem on your side, I guess.

>>  CONFIG_BOARD_INGENIC=y
> 
> This config option does not exist (at least in v6.0-rc). Probably you 
> refer to CONFIG_INGENIC_GENERIC_BOARD.

No, I do not, and yes, it exists.

> As far as I see, this does not choose to build any device tree blob.
> 
> I tried some patch to get the .dtb built, but the resulting kernel 
> does not show any activity.
> 
> If I e.g. switch back from CONFIG_INGENIC_GENERIC_BOARD=y to 
> CONFIG_JZ4780_CI20=y the kernel works.

Because in the first case you build a generic kernel, which does not 
embed any .dtb, and you are responsible for providing the kernel with 
the blob at boot time; while if you build with CONFIG_JZ4780_CI20=y it 
embeds the .dtb inside the kernel.

You can embed the .dtb into the generic kernel at compile-time too, 
have a look at "Kernel type -> Kernel appended dtb support." Not sure 
why you'd want that for a generic kernel, though.

>> 
>>  Then you can support all Ingenic-based boards alongside other MIPS 
>> boards.
> 
> Yes, I know, but why are the MACH_JZ47* not replaced by 
> CONFIG_MACH_INGENIC_GENERIC if they are almost unused or completely 
> removed?

They *are* used.

> BTW: there are also seems to be some board specific CONFIGs in 
> processor specific code (e.g. CONFIG_JZ4780_CI20 in irqchip code).

rgrep CI20 drivers/irqchip/

returns nothing for me.

> So selecting a MACH is not sufficient to get these features.
> 
> All this looks a little fragile and incomplete... Maybe if I find 
> some time (which is unfortunately quite unlikely) I can propose some 
> fixes.

It is not "fragile and incomplete", it works as intended, and it's a 
feature I use often.

-Paul