How to work with Linux-Mips ?

How to work with Linux-Mips ?

[Franck Bui-Huu]

Hi,

First of all, I'm sending this email because I really want to
understand how linux-mips is working with its community. I'm asking
this question because I'm wondering how many times I should resend a
patch before it can be considered for applying.

For example, I posted some patches more than one month ago. I really
think that these patches improve the MIPS specific code and they are
not complex at all:

http://www.linux-mips.org/archives/linux-mips/2006-08/msg00112.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00113.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00114.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00115.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00117.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00118.html

http://www.linux-mips.org/archives/linux-mips/2006-08/msg00196.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00195.html
http://www.linux-mips.org/archives/linux-mips/2006-08/msg00197.html

These patchs have been kindly reviewed and acked by Atsushi Nemoto,
but then no feedback from the MIPS team. I tried to get a status for
MIPS team a couple of time, to know if something was wrong with them
but MIPS people seem to not care about them. They even haven't
bothered to take 10 seconds for replying something like:

  - your patches are broken because...

  - your patches do not respect our MIPS protocol, please resend...

  - Sorry we are very busy, please hold on...

  - OK your patches suck please try to work on ARM chips because MIPS
    is a very closed circle reserved to MIPS gurus.

Another question, is that MIPS tree seems to not care about linux
mainline release process. I actually notice that even Linus do not
pull MIPS tree anymore during the last release candidate cycle. Is
MIPS aware that some of its customers are trying to make stable
releases ? Does the linux-mips team exist to ease life of its
customers to use the linux kernel on MIPS chips or is the purpose of
this team doing only some development for fun ?

Hoping this time I'll get some answers.

		Franck

Re: How to work with Linux-Mips ?

[Ralf Baechle]

On Tue, Sep 26, 2006 at 09:14:07AM +0200, Franck Bui-Huu wrote:

> These patchs have been kindly reviewed and acked by Atsushi Nemoto,
> but then no feedback from the MIPS team. I tried to get a status for
> MIPS team a couple of time, to know if something was wrong with them
> but MIPS people seem to not care about them. They even haven't
> bothered to take 10 seconds for replying something like:
> 
>   - your patches are broken because...
> 
>   - your patches do not respect our MIPS protocol, please resend...
> 
>   - Sorry we are very busy, please hold on...

You should interpret a non-answer as exactly this.

>   - OK your patches suck please try to work on ARM chips because MIPS
>     is a very closed circle reserved to MIPS gurus.
> 
> Another question, is that MIPS tree seems to not care about linux
> mainline release process. I actually notice that even Linus do not
> pull MIPS tree anymore during the last release candidate cycle.

While my employer encourages my linux-mips.org activities at times there
are things that take priority and that has happened before 2.6.18 until
the time where it was too late to submit anything.  Still 2.6.18 from
kernel.org is usable on MIPS.

> Is MIPS aware that some of its customers are trying to make stable
> releases ?

I pass on answering a polemic question.

> Does the linux-mips team exist to ease life of its
> customers to use the linux kernel on MIPS chips or is the purpose of
> this team doing only some development for fun ?

You should draw a line between linux-mips.org and MIPS Technologies.
linux-mips.org isn't a formal organization and not to be considered part
of MTI.  If anything call it a loosely knit group of hackers with
interest in running Linux and anything and all that has a MIPS processors.
There is more to be said I guess you rather want me to return to my
pending patches ;-)

  Ralf

Re: How to work with Linux-Mips ?

[Franck Bui-Huu]

Hi Ralf,

I'm glad you answered...

Ralf Baechle wrote:
> On Tue, Sep 26, 2006 at 09:14:07AM +0200, Franck Bui-Huu wrote:
> 
>> These patchs have been kindly reviewed and acked by Atsushi Nemoto,
>> but then no feedback from the MIPS team. I tried to get a status for
>> MIPS team a couple of time, to know if something was wrong with them
>> but MIPS people seem to not care about them. They even haven't
>> bothered to take 10 seconds for replying something like:
>>
>>   - your patches are broken because...
>>
>>   - your patches do not respect our MIPS protocol, please resend...
>>
>>   - Sorry we are very busy, please hold on...
> 
> You should interpret a non-answer as exactly this.
> 

Well, I usually do this interpretation at first but the patches were
posted more than one month ago and during this while you considered,
reviewed and applied several other patches.

Futhermore replying with a short email saying that you're _very_ busy
and you'll deal with them after 2.6.18 releasing would have been
welcome. And it would have taken less than 1 minute of your time.

>> Does the linux-mips team exist to ease life of its
>> customers to use the linux kernel on MIPS chips or is the purpose of
>> this team doing only some development for fun ?
> 
> You should draw a line between linux-mips.org and MIPS Technologies.
> linux-mips.org isn't a formal organization and not to be considered part
> of MTI.  If anything call it a loosely knit group of hackers with
> interest in running Linux and anything and all that has a MIPS processors.

Thanks for this clarification, which is a bit scaring, I wasn't aware
of this.

> There is more to be said I guess you rather want me to return to my
> pending patches ;-)
> 

yes please go ahead. Just note that this email is not an attack or
whatever, I'm just trying to understand what I'm doing wrong.

		Franck

Re: How to work with Linux-Mips ?

[Peter Popov]

<snip>


> These patchs have been kindly reviewed and acked by Atsushi Nemoto,
> but then no feedback from the MIPS team. I tried to get a status for
> MIPS team a couple of time, to know if something was wrong with them
> but MIPS people seem to not care about them. They even haven't
> bothered to take 10 seconds for replying something like:


>   - your patches are broken because...
> 
>   - your patches do not respect our MIPS protocol, please resend...
> 
>   - Sorry we are very busy, please hold on...
> 
>   - OK your patches suck please try to work on ARM chips because MIPS
>     is a very closed circle reserved to MIPS gurus.

There is no such thing as a 10 sec patch review. Especially when it
comes to patches that touch generic portions of the kernel. Someone has
to very carefully consider what they are doing, any potential risks, how
to test them, etc. Multiply that times the number of people sending
patches plus the new development that someone like Ralf is doing. Try
working full time and being a maintainer yourself for a single board
only over a period of a few years.

> Another question, is that MIPS tree seems to not care about linux
> mainline release process. I actually notice that even Linus do not
> pull MIPS tree anymore during the last release candidate cycle. Is
> MIPS aware that some of its customers are trying to make stable
> releases ? 

MIPS Tech and linux-mips are separate entities. Personally I think MIPS
Tech has done a great service to themselves, their customers, and the
entire Linux MIPS community, by hiring Ralf to do new MIPS development
and properly support new chips coming out.

> Does the linux-mips team exist to ease life of its
> customers to use the linux kernel on MIPS chips or is the purpose of
> this team doing only some development for fun ?

Does it really matter? The developers that are active on the linux-mips
mailing list are not associated with MIPS Tech nor are they paid for
their effort by you or the company you work for. If you want a
"product", you can buy a commercial embedded Linux distro from someone.
If you want to contribute to the linux mips development effort, then
learn to live with any frustrations associated with working in community
trees. And next time, just ask nicely what you can do to get some
feedback on your patches.

Pete

Re: How to work with Linux-Mips ?

[Jonathan Day]

--- Peter Popov <ppopov@embeddedalley.com> wrote:
> There is no such thing as a 10 sec patch review.
> Especially when it
> comes to patches that touch generic portions of the
> kernel.

That is extremely true. One has to only look at the
number of "Brown Paper Bag" releases, over the history
of Linux, to realize that even the intense reviews
that take place are sometimes not enough.

> MIPS Tech and linux-mips are separate entities.
> Personally I think MIPS
> Tech has done a great service to themselves, their
> customers, and the
> entire Linux MIPS community, by hiring Ralf to do
> new MIPS development
> and properly support new chips coming out.

Whilst I agree entirely, I think we need to put a
little more perspective on this. The Linux kernel is
big. Very, very big. By my estimate, it would take an
army of 10,000+ full-time software engineers skilled
in "Extreme Programming" and formal methods to be able
to verify something of the complexity and intricacy of
the Linux kernel within a single year, excluding any
changes made during that time, which will likely
replace so much of the code that the verification
won't tell you much anyway.

If every company and every University involved in
Linux - not just every consortium - were to
contribute, you might be able to amass that kind of
manpower. One full-time coder for the entire of the
MIPS side of the tree is valuable and it's doubtful
any branch could now survive long without at least
that, one person is simply not capable of replacing
ten thousand, no matter how brilliant they are.

This isn't to say MIPS Tech should necessarily throw
in more manpower, although I certainly wouldn't argue
with that. It's that there's simply no realistic way
to get enough manpower together to do code reviews in
the kind of timeframes that people are asking for.

Rather than curse out the absence of something that is
beyond any realistic reach, it might be good for
people to reflect on the sheer magnitude of what we do
have.

I'd also like to throw in one other thought - people
have patches and patch-sets on web pages that are
popular but don't get merged in for ages, or sometimes
at all. I know this as well as anyone can - I ran a
project for a while that did nothing but collect the
really significant patches and patchsets out there and
merge them into one gigantic megapatch. It was about
two-thirds the size of the kernel itself.

Even at that kind of size, there was probably as much
out there that I knew about but omitted as I included,
and very likely ten times as much as all that combined
that I never knew about at all.

Today, with far more developers, far more hardware,
far more gadgets and gizmos, vastly more R&D, and
infinitely more interest in Linux, the ratio of
included code to code "missing, presumed blogged" can
only be worse. The various development teams don't
scale nearly as well or as fast as the Internet.

On that basis, my guess is that there is probably
between 15-20 times as much interesting code out there
that has not been reviewed and included as there is
code in the whole of the Linux kernel as it currently
stands. If, as may well happen, interest goes through
another explosion before 2.8 gets out, then this could
easily double over the next three or four years.

Yes, I wish things moved faster. Yes, I wish there
were fewer problems with some of the boards I use. I
also wish I knew next week's lottery numbers. The odds
of me finding a valid solution to the third seems
infinitely more likely than anyone developing a
perfect, lightning-speed solution to the first two.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

Re: How to work with Linux-Mips ?

[Franck Bui-Huu]

Peter Popov wrote:
> <snip>
> 
> 
>> These patchs have been kindly reviewed and acked by Atsushi Nemoto,
>> but then no feedback from the MIPS team. I tried to get a status for
>> MIPS team a couple of time, to know if something was wrong with them
>> but MIPS people seem to not care about them. They even haven't
>> bothered to take 10 seconds for replying something like:
> 
> 
>>   - your patches are broken because...
>>
>>   - your patches do not respect our MIPS protocol, please resend...
>>
>>   - Sorry we are very busy, please hold on...
>>
>>   - OK your patches suck please try to work on ARM chips because MIPS
>>     is a very closed circle reserved to MIPS gurus.
> 
> There is no such thing as a 10 sec patch review. Especially when it
> comes to patches that touch generic portions of the kernel. Someone has
> to very carefully consider what they are doing, any potential risks, how

Who is claiming that ? I've never said that patch review needs 10 sec to
be done....

> 
> MIPS Tech and linux-mips are separate entities. Personally I think MIPS
> Tech has done a great service to themselves, their customers, and the
> entire Linux MIPS community, by hiring Ralf to do new MIPS development
> and properly support new chips coming out.

Again, none has said the contrary...

		Franck

Re: How to work with Linux-Mips ?

[Franck Bui-Huu]

Hi

Jonathan Day wrote:
> Whilst I agree entirely, I think we need to put a
> little more perspective on this. The Linux kernel is
> big. Very, very big. By my estimate, it would take an
> army of 10,000+ full-time software engineers skilled
> in "Extreme Programming" and formal methods to be able
> to verify something of the complexity and intricacy of
> the Linux kernel within a single year, excluding any
> changes made during that time, which will likely
> replace so much of the code that the verification
> won't tell you much anyway.
> 
> If every company and every University involved in
> Linux - not just every consortium - were to
> contribute, you might be able to amass that kind of
> manpower. One full-time coder for the entire of the
> MIPS side of the tree is valuable and it's doubtful
> any branch could now survive long without at least
> that, one person is simply not capable of replacing
> ten thousand, no matter how brilliant they are.
> 

Well, I'm not sure about that. Now knowing that Ralf is not
maintaining/developing on linux-mips tree full time is a good
information to keep in mind. I think he's able to take care of any
MIPS specific code if he had time. The current number of patches sent
to linux-mips is not so big. Look at how Andrew Morton is maintaining
the mm branch. This branch is far more active and patches included in
this tree are far more complex than those sent to this mailing list.

> This isn't to say MIPS Tech should necessarily throw
> in more manpower, although I certainly wouldn't argue
> with that. It's that there's simply no realistic way
> to get enough manpower together to do code reviews in
> the kind of timeframes that people are asking for.
> 

Well IMHO MIPS Tech may be under-estimating its Linux kernel port. I
think it's a real advantage to sell some hardwares and to give some
softwares that is already running on that hardware, specially if this
software is a full operating system like Linux. But if MIPS' customers
don't know on how fast this software will involve, will be fixed, it's
a bit scaring to rely on that software.

It would be interesting to know how well Linux is supported on the
different arches keeping in mind the market that they are targeting
(embedded, huge server...).

		Franck