Linux-next testing

Linux-next testing

[Stoyan Gaydarov]

I have a box that I would like to devote to test the linux-next
kernels. I am however not very experienced with the whole process. I
installed a clean Ubuntu server on it because thats what I had at
home, I know how to compile the kernel, and install(make; make
modules; make install) I also know how to log the build so that if
theres any errors i can have a record of what went wrong. What I need
help is knowing what to test and getting reports back to you. I know I
can do a 'git clone "path to linux-next git tree" ' and go into that
and use randconfig then install and reboot, but I was wondering if
there was some other better way to do it because the clone command
only works in an empty directory. I think fetch or something like that
might work better. Anyway I also wanted to know which configuration
needed the most testing or if I should just stick with the
randconfig(btw this is an x86 box). The other question is about
reporting failures, which i can almost guarantee will happen, one way
is to keep a vanilla kernel installed so that i can aways go back to
it and report the log files but if its a oops or a BUG() call i am not
sure they get reported there... The last question I have is regarding
the load on the box, this is not a very new box so I cant run too many
things at once but I would like to know if there is anything I should
be running on the box while it is running that way it can stress the
system a little and try to break it in a way that can be useful.
Looking back at what I wrote, those are not quite questions but I
think you get the idea, I just need some information on what to do so
that I can be of use to the people that are developing with
linux-next.

-Stoyan G

Re: Linux-next testing

[Kyle Spaans]

On Wed, Jul 16, 2008 at 06:27:27AM -0500, Stoyan Gaydarov wrote:
> I know I
> can do a 'git clone "path to linux-next git tree" ' and go into that
> and use randconfig then install and reboot, but I was wondering if
> there was some other better way to do it because the clone command
> only works in an empty directory. I think fetch or something like that
> might work better.
Have a look in the archives for this list. Every time Stephen announces a new linux-next release, he includes a blurb at the bottom of the email that will tell you how you can get the linux-next tree, besides just cloning it anew each day. (I found that blurb helpful even on top of the wiki at http://linux.f-seidel.de/linux-next/pmwiki/)

> The last question I have is ... stress the system a little and try to
>  break it in a way that can be useful.
Look around on the linux-next wiki, and you'll see that other people are running automated testing. Have a look at how they are doing it. Also look around on the kernelnewbies.org website, they have tips for how to stress-test your system. (bonnie++ as a filesystem benchmark springs immediately to mind, prime95 as a way to generate cpu load, etc...). I'm sure a quick internet search for "linux stress test" will also turn up useful results.

I'm largely doing the same thing as you, so I'll be interested to see the answers to your other questions.

gl & hf

Re: Linux-next testing

[Stoyan Gaydarov]

On Wed, Jul 16, 2008 at 9:43 AM, Kyle Spaans
<kspaans@student.math.uwaterloo.ca> wrote:
> On Wed, Jul 16, 2008 at 06:27:27AM -0500, Stoyan Gaydarov wrote:
>> I know I
>> can do a 'git clone "path to linux-next git tree" ' and go into that
>> and use randconfig then install and reboot, but I was wondering if
>> there was some other better way to do it because the clone command
>> only works in an empty directory. I think fetch or something like that
>> might work better.
> Have a look in the archives for this list. Every time Stephen announces a new linux-next release, he includes a blurb at the bottom of the email that will tell you how you can get the linux-next tree, besides just cloning it anew each day. (I found that blurb helpful even on top of the wiki at http://linux.f-seidel.de/linux-next/pmwiki/)
>
>> The last question I have is ... stress the system a little and try to
>>  break it in a way that can be useful.
> Look around on the linux-next wiki, and you'll see that other people are running automated testing. Have a look at how they are doing it. Also look around on the kernelnewbies.org website, they have tips for how to stress-test your system. (bonnie++ as a filesystem benchmark springs immediately to mind, prime95 as a way to generate cpu load, etc...). I'm sure a quick internet search for "linux stress test" will also turn up useful results.
>
> I'm largely doing the same thing as you, so I'll be interested to see the answers to your other questions.

I was actually thinking about doing something with uml, where i could
build several kernels with ARCH=um SUBARCH=** that way I could boot
them without having to restart the computer and I could also log all
of the output that was generated a lot easier, and the other thing is
I can have multiple kernels booted at the same time. This would make
the stress tests a lot more effective if theres even less resources
per uml kernel running. Any suggestions on how to do this effectively
and correctly(since i am not an expert in any of these topics) would
be greatly appreciated.

-Stoyan G

>
> gl & hf
> --
> To unsubscribe from this list: send the line "unsubscribe linux-next" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>