linux-kernel - Re: [PATCH] ARC: Enable machine_desc->init_per

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: <eb9111fb-7530-98d9-6800-8413ad90360c@synopsys.com>
Date:   Wed, 20 Jun 2018 16:11:27 -0700
From:   Vineet Gupta <Vineet.Gupta1@...opsys.com>
To:     Alexey Brodkin <Alexey.Brodkin@...opsys.com>
CC:     "wbx@...ibc-ng.org" <wbx@...ibc-ng.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-snps-arc@...ts.infradead.org" 
        <linux-snps-arc@...ts.infradead.org>
Subject: Re: [PATCH] ARC: Enable machine_desc->init_per_cpu for non-SMP
 configs

On 06/14/2018 03:26 PM, Alexey Brodkin wrote:
> Hi Vineet,
>
> On Thu, 2017-11-30 at 09:46 -0800, Vineet Gupta wrote:
>> On 11/29/2017 12:21 AM, Alexey Brodkin wrote:
>>> As of today we assumed that "machine_desc->init_per_cpu" calls
>>> are only usable on SMP systems when we want to run some piece of
>>> code on early boot for each and every core, I guess assumption was
>>> we have "machine_desc->init_early" for single-core cases where
>>> the one and only master core can do all the things.
>>>
>>> But it turned out for platforms which might be both UP and SMP it
>>> might be benificial to use "init_per_cpu" for both UP and SMP cases
>>> with which we achieve 2 things simultaneously:
>>>   1) Exactly the same one code will be used for UP&SMP for
>>>      things required to be done on each an every core regardless if it's
>>>      a master and the only core in UP system or any other slave core in SMP
>>>      setup.
>>>   1) There will be no "ifdef CONFIG_SMP" around "init_per_cpu".
>>>
>> Seems fine to me. However this needs to go with the actual platform change which 
>> needs it.
> Well for example this might get in the way of building kernel for HSDK with
> CONFIG_SMP disabled which is IMHO quite valid case in terms of testing
> code-base compiled with no CONFIG_SMP 

Fair enough - although world is moving away from UP.


> (something Waldemar was up to).

Really, why does he care if we are running SMP kernel - oh his simulation time
goes up. But honestly he should try and use the SMP kernel for testing as that
might uncover any system wide / SMP specific bugs in testing uclibc as well !

> Is it a strong enough reason for that patch to be applied?

Well per my original reply I didn't have any objections for this per-se and wanted
to see the platform patch. Seems we don't need platform patch as it is not under
#ifdef in hsdk/platform.c

I'll queue this up !

-Vineet