All of lore.kernel.org
 help / color / mirror / Atom feed
From: Piotr Hosowicz <piotr@hosowicz.com>
To: Randy Dunlap <rdunlap@xenotime.net>
Cc: linux-kbuild@vger.kernel.org
Subject: Re: Maybe my frequent kernel building could be of some help to the community? And how to build?
Date: Thu, 25 Mar 2010 19:35:21 +0100	[thread overview]
Message-ID: <4BABACE9.8000405@example.com> (raw)
In-Reply-To: <4BABA33D.1080903@xenotime.net>

Randy Dunlap wrote:
> On 03/25/10 10:48, Piotr Hosowicz wrote:
>> Randy Dunlap wrote:
>>> On 03/25/10 10:37, Piotr Hosowicz wrote:
>>>> Randy Dunlap wrote:
>>>>> On 03/18/10 08:40, Randy Dunlap wrote:
>>>>>> On Tue, 16 Mar 2010 12:24:35 -0700 Randy Dunlap wrote:
>>>>>>
>>>>>>> On 03/16/10 12:02, Piotr Hosowicz wrote:
>>>>>>>> Hello,
>>>>>>>>
>>>>>>>> I own quite strong PC, 4 core Pentium. I recently started to
>>>>>>>> build my
>>>>>>>> kernel frequently, whenever I see a new patch on kernel.org I build
>>>>>>>> again. Bothers me that I do it only for doing it. Maybe it could
>>>>>>>> be of
>>>>>>>> some help to the community? And if so, what would be more
>>>>>>>> appreciated -
>>>>>>>> linux-next patched kernel or gitN patched kernel? And how should I
>>>>>>>> configure the kernel for this purpose? I am afraid of putting all
>>>>>>>> the
>>>>>>>> options on, yesterday I tried to put everything on as a module
>>>>>>>> plus I
>>>>>>>> tinkered with some other options that seemed innocent. The kernel
>>>>>>>> was
>>>>>>>> built but could not boot, saying that there was a problem with the
>>>>>>>> root
>>>>>>>> device, it could not recoginze /dev/sda2, which is my correct root
>>>>>>>> device. I did not investigate much and reverted to working kernel.
>>>>>>>>
>>>>>>>> Thanks in advance for your remarks.
>>>>>>> Many linux-next builds are done daily: see
>>>>>>>   http://kisskb.ellerman.id.au/kisskb/branch/9/
>>>>>>>
>>>>>>> and then I do 25 x86_32 randconfigs and 25 x64_64 randconfigs
>>>>>>> daily and fix most build errors that I find in them.
>>>>>>> (I am also using a quad-core PC with a cron job that finds the
>>>>>>> next day's linux-next, downloads it, and does 50 builds.)
>>>>>>>
>>>>>>> You are welcome to do the same or test mainline (with daily
>>>>>>> builds) and/or Andrew Morton's mmotm patchset.
>>>>>> One thing that you could do that no one focuses on is building
>>>>>> with some kconfig symbols disabled (ones that are typically enabled),
>>>>>> such as CONFIG_SMP=n, CONFIG_SYSFS=n, CONFIG_PROC_FS=n, CONFIG_PM=n,
>>>>>> CONFIG_PCI=n, CONFIG_BLOCK=n, CONFIG_NET=n, CONFIG_INET=n (but latter
>>>>>> with CONFIG_NET=y), CONFIG_HOTPLUG=n.  Not all of these at the same
>>>>>> time,
>>>>>> just various/random combinations of them.
>>>>> and CONFIG_SYSRQ=n  (just found some errors due to this one)
>>>> You mean CONFIG_SYSRQ=n exactly and mandatory in every config or this
>>>> was just an example?
>>> Sorry about that.  Just make it random (sometimes disabled).
>> Ok.
>>
>> The problem I face and want to overcome now is that not all options can
>> be random, some options require to be not y or no a tall, but a string
>> or a number. So my plan is that first thing to be done is to do make
>> defconfig and put it as a template to be used by my scripts. Then only
>> the options that you enumerated earlier will be randomized. Is this
>> correct?
> 
> Yes, that's one way to approach it.
> 
> Another way (similar) would be to use a combination of "make randconfig"
> with KCONFIG_ALLCONFIG.  See Documentation/kbuild/kconfig.txt for info
> on KCONFIG_ALLCONFIG.  Basically it allows you to tell kconfig values for
> a subset of config symbols to be used.

Ok, but at present is not working properly, as expected. The scripts now 
are like this, I returned to randomizing with bash:

--- try-build.sh

#!/bin/bash

RCPT=phz@localhost
CPUCNT=$(cat /proc/cpuinfo | grep ^processor | wc -l)

function tidy(){
   rm ../results/.config 2>/dev/null
   rm ../results/build.log 2>/dev/null
   rm ../BUILDING-NOW 2>/dev/null
}

cd ~/kernel/linux
if test -f ../BUILDING-NOW ; then
   exit 0
else
   touch ../BUILDING-NOW
   make mrproper
   cat ../config.all | ../randomize-conf2.sh > ./.config
   make oldconfig
   cp .config ../results/
   DT=$(date +%y%m%d-%H%M%S)	
   make -j $CPUCNT vmlinux 2>&1 | tee ../results/build.log
   if test $? -ne 0 ; then
     (cat ../results/.config && cat ../results/build.log) | mail -s 
'Build failed at kernel phase' $RCPT
     mv ../results/.config ../results/.config.kf.$DT
     mv ../results/build.log ../results/build.log.$DT
     tidy
     exit 0
   fi
   make -j $CPUCNT modules 2>&1 | tee -a ../results/build.log
   if test $? -ne 0 ; then
     (cat ../results/.config && cat ../results/build.log) | mail -s 
'Build failed at modules phase' $RCPT
     mv ../results/.config ../results/.config.mf.$DT
     mv ../results/build.log ../results/build.log.$DT
     tidy
     exit 0
   fi
   cp ../results/build.log ./
   rm BUILDING-NOW
   echo OK | mail -s 'ok' $RCPT
   tidy

fi


--- randomize-conf2.sh:

#!/bin/bash

NETENABLED=0
function enable_or_not(){
   if test $RANDOM -lt 16383 ; then
     if test $1=CONFIG_NET ; then
       NETENABLED=1
       echo $1=y
       return
     fi
     if test $1=CONFIG_INET && test $NETENABLED -eq 1 ; then
       if test $RANDOM -lt 16383 ; then
         echo $1=y
       else
         echo '# '$1' is not set'
       fi
       return
     fi
   else
     echo '# '$1' is not set'
   fi
}

cat ../config.all | grep -v CONFIG_SMP | grep -v CONFIG_SYSFS | grep -v 
CONFIG_PROC_FS | grep -v CONFIG_PM | grep -v CONFIG_PCI | grep -v 
CONFIG_BLOCK | grep -v CONFIG_NET | grep -v CONFIG_INET | grep -v 
CONFIG_HOTPLUG > ./cfg.tmp

(
enable_or_not CONFIG_SMP
enable_or_not CONFIG_SYSFS
enable_or_not CONFIG_PROC_FS
enable_or_not CONFIG_PM
enable_or_not CONFIG_PCI
enable_or_not CONFIG_BLOCK
enable_or_not CONFIG_NET
enable_or_not CONFIG_INET
enable_or_not CONFIG_HOTPLUG
)
cat ./cfg.tmp

I launch try-build.sh, but it still asks me about unspecified options, 
my config.all is produced by make defconfig or make oldconfig, dosen't 
matter still.

Do you see what the problem is?


-- 
Z cyklu "Uroki demokracji", czyli pytania i odpowiedzi w teledurniejach:
- Ile kół ma trójkołowiec?
- Dwa.
NP: Miles Davis and John Coltrane - Two Bass Hit
NB: 2.6.34-rc2-nx-20100325-next-20100325+

  reply	other threads:[~2010-03-25 18:35 UTC|newest]

Thread overview: 47+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-03-16 19:02 Maybe my frequent kernel building could be of some help to the community? And how to build? Piotr Hosowicz
2010-03-16 19:24 ` Randy Dunlap
2010-03-16 19:33   ` Piotr Hosowicz
2010-03-16 19:42     ` Randy Dunlap
2010-03-17  3:24   ` Piotr Hosowicz
2010-03-17 15:07     ` Randy Dunlap
     [not found]       ` <4BA0F8DE.1050006@example.com>
2010-03-17 15:54         ` Randy Dunlap
2010-03-17 16:03           ` Piotr Hosowicz
2010-03-17 17:49           ` Piotr Hosowicz
2010-03-17 17:53             ` Randy Dunlap
2010-03-18 15:40   ` Randy Dunlap
2010-03-22  3:31     ` Piotr Hosowicz
2010-03-22  3:48       ` Randy Dunlap
2010-03-22 12:42         ` Piotr Hosowicz
2010-03-22 12:53           ` Piotr Hosowicz
2010-03-24  1:12             ` Piotr Hosowicz
2010-03-24 21:43               ` Randy Dunlap
2010-03-25 11:38                 ` Piotr Hosowicz
2010-03-25 17:29     ` Randy Dunlap
2010-03-25 17:37       ` Piotr Hosowicz
2010-03-25 17:40         ` Randy Dunlap
2010-03-25 17:48           ` Piotr Hosowicz
2010-03-25 17:54             ` Randy Dunlap
2010-03-25 18:35               ` Piotr Hosowicz [this message]
2010-03-25 22:04                 ` Randy Dunlap
2010-03-25 22:15                   ` Piotr Hosowicz
2010-03-26 16:46                     ` Piotr Hosowicz
2010-03-27 16:37                       ` Piotr Hosowicz
2010-03-27 16:39                         ` Piotr Hosowicz
2010-04-01 16:02       ` Randy Dunlap
2010-04-01 17:24         ` Piotr Hosowicz
2010-04-01 17:44           ` Piotr Hosowicz
2010-03-29 20:16   ` Is root kernel parameter expected to be the first one or something? Piotr Hosowicz
2010-03-29 20:38     ` Randy Dunlap
2010-03-29 20:41       ` Piotr Hosowicz
2010-03-29 21:33         ` Randy Dunlap
2010-03-29 22:32           ` Piotr Hosowicz
2010-03-29 22:39             ` Randy Dunlap
2010-03-29 22:44               ` Piotr Hosowicz
2010-03-29 22:48                 ` Randy Dunlap
2010-03-29 22:55                   ` Piotr Hosowicz
2010-03-29 23:06                     ` Randy Dunlap
2010-03-29 23:44                       ` Piotr Hosowicz
2010-03-30 16:37                   ` Piotr Hosowicz
2010-03-30 18:03                     ` Piotr Hosowicz
2010-03-30 19:00                       ` Piotr Hosowicz
2010-03-31  9:15                         ` Piotr Hosowicz

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4BABACE9.8000405@example.com \
    --to=piotr@hosowicz.com \
    --cc=linux-kbuild@vger.kernel.org \
    --cc=rdunlap@xenotime.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.