[9fans] apropos of the glendix post

[9fans] apropos of the glendix post

[ron minnich]

I just stumbled across a talk I gave in 2002:

Here's one slide:

How to fix this (2): 9sys
We are building a Plan 9 system call set into Linux
It is only 38 calls
Three steps
Initial support as ioctl's from a device (/dev/9sys)
Direct integration into Linux system call table
Remove non-Plan 9 system calls (ca. 235 of them)
Result should be Linux portability with Plan 9 architecture
Greatly pared-down footprint
Early code is working now

Note that the system calls back then were 235 in number, are 300 now.
Note that one of the steps was "remove

good luck with glendix, it will be fun if it works. I stopped work on
9sys when lucent fixed the license.

ron

Re: [9fans] apropos of the glendix post

[Uriel]

Interesting. Where is the source for 9sys?

Peace

uriel


On Wed, Sep 3, 2008 at 1:09 AM, ron minnich <rminnich@gmail.com> wrote:
> I just stumbled across a talk I gave in 2002:
>
> Here's one slide:
>
> How to fix this (2): 9sys
> We are building a Plan 9 system call set into Linux
> It is only 38 calls
> Three steps
> Initial support as ioctl's from a device (/dev/9sys)
> Direct integration into Linux system call table
> Remove non-Plan 9 system calls (ca. 235 of them)
> Result should be Linux portability with Plan 9 architecture
> Greatly pared-down footprint
> Early code is working now
>
> Note that the system calls back then were 235 in number, are 300 now.
> Note that one of the steps was "remove
>
> good luck with glendix, it will be fun if it works. I stopped work on
> 9sys when lucent fixed the license.
>
> ron
>
>

Re: [9fans] apropos of the glendix post

[ron minnich]

On Tue, Sep 2, 2008 at 4:22 PM, Uriel <uriel99@gmail.com> wrote:
> Interesting. Where is the source for 9sys?
>

It's not worth much.

I've got it if anyone wants it but I would bet glendix is further
along and much better.

ron

Re: [9fans] apropos of the glendix post

[Skip Tavakkolian]

other than trying every itiration possible (sum over histories of
software?), i'm not getting the reason why glendix is a good idea.

Re: [9fans] apropos of the glendix post

[ron minnich]

On Tue, Sep 2, 2008 at 6:38 PM, Skip Tavakkolian <9nut@9netics.com> wrote:
> other than trying every itiration possible (sum over histories of
> software?), i'm not getting the reason why glendix is a good idea.
>

for me, it's the same thing over and over again. drivers.

ron

Re: [9fans] apropos of the glendix post

[erik quanstrom]

> On Tue, Sep 2, 2008 at 6:38 PM, Skip Tavakkolian <9nut@9netics.com> wrote:
> > other than trying every itiration possible (sum over histories of
> > software?), i'm not getting the reason why glendix is a good idea.
> >
>
> for me, it's the same thing over and over again. drivers.

what drivers are the problem?

- erik

Re: [9fans] apropos of the glendix post

[Anant Narayanan]

On 03-Sep-08, at 3:02 PM, erik quanstrom wrote:
>> On Tue, Sep 2, 2008 at 6:38 PM, Skip Tavakkolian <9nut@9netics.com>
>> wrote:
>>> other than trying every itiration possible (sum over histories of
>>> software?), i'm not getting the reason why glendix is a good idea.
>>>
>>
>> for me, it's the same thing over and over again. drivers.
>
> what drivers are the problem?

Any ethernet or video card that Plan 9 doesn't support (there are
plenty).

--
Anant

Re: [9fans] apropos of the glendix post

[erik quanstrom]

>>> for me, it's the same thing over and over again. drivers.
>>
>> what drivers are the problem?
>
> Any ethernet or video card that Plan 9 doesn't support (there are
> plenty).

surely you don't own *all* the unsupported cards?

- erik

Re: [9fans] apropos of the glendix post

[lucio]

> surely you don't own *all* the unsupported cards?

The growing integration makes you own at least one of each :-(

I got the impression that Ron's particular bugbear was the integrated
wireless adapter.

++L

Re: [9fans] apropos of the glendix post

[Anant Narayanan]

On 03-Sep-08, at 7:45 PM, erik quanstrom wrote:
>>>> for me, it's the same thing over and over again. drivers.
>>>
>>> what drivers are the problem?
>>
>> Any ethernet or video card that Plan 9 doesn't support (there are
>> plenty).
>
> surely you don't own *all* the unsupported cards?

No, but I use a Macbook Pro, and I have friends who would like to run
Plan 9 but can't because of the hardware they own. Think of Glendix as
a stop-gap solution until they get new hardware or until more drivers
are written for Plan 9 :-)

--
Anant

Re: [9fans] apropos of the glendix post

[Latchesar Ionkov]

there are more people working on stop-gap solutions than drivers so I
wonder how long we'll have to wait for the drivers to be written :)

On Wed, Sep 3, 2008 at 12:17 PM, Anant Narayanan <anant@kix.in> wrote:
> On 03-Sep-08, at 7:45 PM, erik quanstrom wrote:
>>>>>
>>>>> for me, it's the same thing over and over again. drivers.
>>>>
>>>> what drivers are the problem?
>>>
>>> Any ethernet or video card that Plan 9 doesn't support (there are
>>> plenty).
>>
>> surely you don't own *all* the unsupported cards?
>
> No, but I use a Macbook Pro, and I have friends who would like to run Plan 9
> but can't because of the hardware they own. Think of Glendix as a stop-gap
> solution until they get new hardware or until more drivers are written for
> Plan 9 :-)
>
> --
> Anant
>
>
>

Re: [9fans] apropos of the glendix post

[Latchesar Ionkov]

wireless network cards.

On Wed, Sep 3, 2008 at 7:02 AM, erik quanstrom <quanstro@coraid.com> wrote:
>> On Tue, Sep 2, 2008 at 6:38 PM, Skip Tavakkolian <9nut@9netics.com> wrote:
>> > other than trying every itiration possible (sum over histories of
>> > software?), i'm not getting the reason why glendix is a good idea.
>> >
>>
>> for me, it's the same thing over and over again. drivers.
>
> what drivers are the problem?
>
> - erik
>
>

Re: [9fans] apropos of the glendix post

[ron minnich]

On Wed, Sep 3, 2008 at 2:12 PM, Latchesar Ionkov <lucho@ionkov.net> wrote:
> there are more people working on stop-gap solutions than drivers so I
> wonder how long we'll have to wait for the drivers to be written :)

given the increasing use of binary blobs, maybe forever.

ron

Re: [9fans] apropos of the glendix post

[Geoffrey Avila]

> On Wed, Sep 3, 2008 at 2:12 PM, Latchesar Ionkov <lucho@ionkov.net> wrote:
>> there are more people working on stop-gap solutions than drivers so I
>> wonder how long we'll have to wait for the drivers to be written :)
>
> given the increasing use of binary blobs, maybe forever.
>
> ron
>

Perhaps integrating support for vendor binary-only drivers would be
worthwhile...woulden't this also provide a good excuse to do ioctl as
well?*


*(kidding...kidding...)

Re: [9fans] apropos of the glendix post

[Latchesar Ionkov]

if linux can use binary blobs, why can't plan9 do it too?

On Wed, Sep 3, 2008 at 4:03 PM, ron minnich <rminnich@gmail.com> wrote:
> On Wed, Sep 3, 2008 at 2:12 PM, Latchesar Ionkov <lucho@ionkov.net> wrote:
>> there are more people working on stop-gap solutions than drivers so I
>> wonder how long we'll have to wait for the drivers to be written :)
>
> given the increasing use of binary blobs, maybe forever.
>
> ron
>
>

Re: [9fans] apropos of the glendix post

[lucio]

> if linux can use binary blobs, why can't plan9 do it too?

I think the BLOBs are platform specific, but I may be mistaken.  If
I'm right, there's no way that we'd get any momentum to turn this
around.

But your earlier comment is extremely valid, too few people are
working on drivers.  Without wishing to fan Discordia's flames, it is
true that the 9fans community does not seem focused on solving
problems, but rather on the academic value of the ease in which
problems could be solved if one followed the 9 True Ways :-)

That said, how do we mobilise the community to focus on useful
drivers?  I suppose we start with Ron's wish list, then we explore
Russ' partially complete postings (i386 emulation, Centrino drivers,
I'm sure I've forgotten many more) and thirdly we post a list of
willing contributors, possibly split into code writers and advisors.

Any takers?

++L

Re: [9fans] apropos of the glendix post

[hiro]

Good luck.

On Thu, Sep 4, 2008 at 8:47 AM,  <lucio@proxima.alt.za> wrote:
>> if linux can use binary blobs, why can't plan9 do it too?
>
> I think the BLOBs are platform specific, but I may be mistaken.  If
> I'm right, there's no way that we'd get any momentum to turn this
> around.
>
> But your earlier comment is extremely valid, too few people are
> working on drivers.  Without wishing to fan Discordia's flames, it is
> true that the 9fans community does not seem focused on solving
> problems, but rather on the academic value of the ease in which
> problems could be solved if one followed the 9 True Ways :-)
>
> That said, how do we mobilise the community to focus on useful
> drivers?  I suppose we start with Ron's wish list, then we explore
> Russ' partially complete postings (i386 emulation, Centrino drivers,
> I'm sure I've forgotten many more) and thirdly we post a list of
> willing contributors, possibly split into code writers and advisors.
>
> Any takers?
>
> ++L
>
>
>

Re: [9fans] apropos of the glendix post

[Michaelian Ennis]

On Thu, Sep 4, 2008 at 1:47 AM,  <lucio@proxima.alt.za> wrote:
>> if linux can use binary blobs, why can't plan9 do it too?
>
> I think the BLOBs are platform specific, but I may be mistaken.  If
> I'm right, there's no way that we'd get any momentum to turn this
> around.

Sometimes the binary bits are loaded into the cards, maybe most often,
and therefore can be done in plan9 as well.  Depending on the vendor
you may need permission from them to distribute the binary part.

> problems, but rather on the academic value of the ease in which
> problems could be solved if one followed the 9 True Ways :-)

Or maybe they have already written the drivers they need, contributed
them to the community, and are busy developing solutions based on
Plan9.

> That said, how do we mobilise the community to focus on useful
> drivers?  I suppose we start with Ron's wish list, then we explore

I suspect that some of these drivers are missing still because the
hardware is not available to the developers who can write them.

> Russ' partially complete postings (i386 emulation, Centrino drivers,
> I'm sure I've forgotten many more) and thirdly we post a list of
> willing contributors, possibly split into code writers and advisors.

I think the one of the BSD projects, OpenBSD perhaps, used to accept
hardware donations to this end.  If we had a wayto get the hardware
and the technical documentation in the same place as the developers we
could get more hardware supported.

Ian

Re: [9fans] apropos of the glendix post

[Tim Wiess]

> That said, how do we mobilise the community to focus on useful
> drivers?  I suppose we start with Ron's wish list, then we explore
> Russ' partially complete postings (i386 emulation, Centrino drivers,
> I'm sure I've forgotten many more) and thirdly we post a list of
> willing contributors, possibly split into code writers and advisors.

Are you serious?
I have a crazy idea: how about you actually write one?

Re: [9fans] apropos of the glendix post

[erik quanstrom]

> Sometimes the binary bits are loaded into the cards, maybe most often,
> and therefore can be done in plan9 as well.  Depending on the vendor
> you may need permission from them to distribute the binary part.

cf. etherm10g.c.  the myricom driver loads a metric ton of firmware into
the card.

> I suspect that some of these drivers are missing still because the
> hardware is not available to the developers who can write them.
>
[...]
>
> I think the one of the BSD projects, OpenBSD perhaps, used to accept
> hardware donations to this end.  If we had a wayto get the hardware
> and the technical documentation in the same place as the developers we
> could get more hardware supported.

good point.  there's plenty of hardware out there with full
documentation that's cheep.  ati has fully documented the 2d
in newer cards, for example.  (a project i never seem to get to.)
intel have done the same.  there are also a number of cards that
aren't really documented but seem straightforward anway — like
the old nvidia propritary sata interface.  if you have such a board
and need >2 drives, this might be a fun and easy project.

writing drivers is an interesting exercize.  on the one hand, small
errors will get you quickly because they typically run millions of
times per second.  this isn't all bad because one finds errors quite
quickly. :-)  on the other hand, they typically follow a
small, constrained interface so it's easy to know when you're done
and you don't need to design an interface.

- erik

Re: [9fans] apropos of the glendix post

[dave.l]

This gets punted around every few months and nothing happens.

I've done some basic information-gathering but got no further for the
usual reasons,
so, in an effort to stimulate some inertia,
here's a small suggestion ...

Is there anyone out there who's gonna be in Volos and wants to
participate in a driver-writing BOF?

DaveL

On 4 Sep 2008, at 06:47, lucio@proxima.alt.za wrote:

>> if linux can use binary blobs, why can't plan9 do it too?
>
> I think the BLOBs are platform specific, but I may be mistaken.  If
> I'm right, there's no way that we'd get any momentum to turn this
> around.
>
> But your earlier comment is extremely valid, too few people are
> working on drivers.  Without wishing to fan Discordia's flames, it is
> true that the 9fans community does not seem focused on solving
> problems, but rather on the academic value of the ease in which
> problems could be solved if one followed the 9 True Ways :-)
>
> That said, how do we mobilise the community to focus on useful
> drivers?  I suppose we start with Ron's wish list, then we explore
> Russ' partially complete postings (i386 emulation, Centrino drivers,
> I'm sure I've forgotten many more) and thirdly we post a list of
> willing contributors, possibly split into code writers and advisors.
>
> Any takers?
>
> ++L
>
>

Re: [9fans] apropos of the glendix post

[ron minnich]

On Thu, Sep 4, 2008 at 4:03 PM,  <dave.l@mac.com> wrote:

> Is there anyone out there who's gonna be in Volos and wants to participate
> in a driver-writing BOF?
>

I'm not gonna make it, but it's an interesting idea.

What would be more useful, possibly, is a tutorial on how to write a
driver, with a few case studies.

ron

Re: [9fans] apropos of the glendix post

[Uriel]

Sounds like a really cool idea, maybe somebody with more experience in
the subject wants to give a short introduction, and then people can
start to work on their favorite wanted driver (alone or in groups)
while the 'expert'(s) in charge of the session can answer questions
and help whoever gets stuck.

Anyone from the Program Committee wants to comment on this? Richard?

Peace

uriel

P.S.: Maybe to make things easier, it would be helpful to setup a
public plan9 server people can both drawterm/cpu to and boot from
their custom kernels for testing purposes, that way even not-fully
functional laptops could be used for development.

On Fri, Sep 5, 2008 at 1:03 AM,  <dave.l@mac.com> wrote:
> This gets punted around every few months and nothing happens.
>
> I've done some basic information-gathering but got no further for the usual
> reasons,
> so, in an effort to stimulate some inertia,
> here's a small suggestion ...
>
> Is there anyone out there who's gonna be in Volos and wants to participate
> in a driver-writing BOF?
>
> DaveL
>
> On 4 Sep 2008, at 06:47, lucio@proxima.alt.za wrote:
>
>>> if linux can use binary blobs, why can't plan9 do it too?
>>
>> I think the BLOBs are platform specific, but I may be mistaken.  If
>> I'm right, there's no way that we'd get any momentum to turn this
>> around.
>>
>> But your earlier comment is extremely valid, too few people are
>> working on drivers.  Without wishing to fan Discordia's flames, it is
>> true that the 9fans community does not seem focused on solving
>> problems, but rather on the academic value of the ease in which
>> problems could be solved if one followed the 9 True Ways :-)
>>
>> That said, how do we mobilise the community to focus on useful
>> drivers?  I suppose we start with Ron's wish list, then we explore
>> Russ' partially complete postings (i386 emulation, Centrino drivers,
>> I'm sure I've forgotten many more) and thirdly we post a list of
>> willing contributors, possibly split into code writers and advisors.
>>
>> Any takers?
>>
>> ++L
>>
>>
>
>
>

Re: [9fans] apropos of the glendix post

[Uriel]

Maybe similar sessions for 'file server writing' or 'limbo
programming' might be nice for those not familiar with those subjects
(and thanks to 9vx and hosted the required setups would allow people
with non-plan9 laptops to participate).

Peace

uriel

On Fri, Sep 5, 2008 at 2:41 AM, Uriel <uriel99@gmail.com> wrote:
> Sounds like a really cool idea, maybe somebody with more experience in
> the subject wants to give a short introduction, and then people can
> start to work on their favorite wanted driver (alone or in groups)
> while the 'expert'(s) in charge of the session can answer questions
> and help whoever gets stuck.
>
> Anyone from the Program Committee wants to comment on this? Richard?
>
> Peace
>
> uriel
>
> P.S.: Maybe to make things easier, it would be helpful to setup a
> public plan9 server people can both drawterm/cpu to and boot from
> their custom kernels for testing purposes, that way even not-fully
> functional laptops could be used for development.
>
> On Fri, Sep 5, 2008 at 1:03 AM,  <dave.l@mac.com> wrote:
>> This gets punted around every few months and nothing happens.
>>
>> I've done some basic information-gathering but got no further for the usual
>> reasons,
>> so, in an effort to stimulate some inertia,
>> here's a small suggestion ...
>>
>> Is there anyone out there who's gonna be in Volos and wants to participate
>> in a driver-writing BOF?
>>
>> DaveL
>>
>> On 4 Sep 2008, at 06:47, lucio@proxima.alt.za wrote:
>>
>>>> if linux can use binary blobs, why can't plan9 do it too?
>>>
>>> I think the BLOBs are platform specific, but I may be mistaken.  If
>>> I'm right, there's no way that we'd get any momentum to turn this
>>> around.
>>>
>>> But your earlier comment is extremely valid, too few people are
>>> working on drivers.  Without wishing to fan Discordia's flames, it is
>>> true that the 9fans community does not seem focused on solving
>>> problems, but rather on the academic value of the ease in which
>>> problems could be solved if one followed the 9 True Ways :-)
>>>
>>> That said, how do we mobilise the community to focus on useful
>>> drivers?  I suppose we start with Ron's wish list, then we explore
>>> Russ' partially complete postings (i386 emulation, Centrino drivers,
>>> I'm sure I've forgotten many more) and thirdly we post a list of
>>> willing contributors, possibly split into code writers and advisors.
>>>
>>> Any takers?
>>>
>>> ++L
>>>
>>>
>>
>>
>>
>

Re: [9fans] apropos of the glendix post

[ron minnich]

Those of us who can't make it could volunteer to be on irc during the
sessions, modulo timezone issues (that's what coffee is for).

ron

Re: [9fans] apropos of the glendix post

[lucio]

> Are you serious?
> I have a crazy idea: how about you actually write one?

I did, once.  At least, I extended the PCNet driver for a particular
model no one else had catered for.  There were a few inconsistencies
with expectations, so it never made prime time.  Since then, the
memory model changed and I couldn't quite figure out the incantations
to make the driver fit.  And the host that needed has long lost its
proprietary power supply.

Next question?

++L

Re: [9fans] apropos of the glendix post

[lucio]

> I think the one of the BSD projects, OpenBSD perhaps, used to accept
> hardware donations to this end.  If we had a wayto get the hardware
> and the technical documentation in the same place as the developers we
> could get more hardware supported.

 I believe this is a small factor, but also not an insignificant one.
 Perhaps, if we could show some seriousness, donations would be more
 forthcoming...

++L

Re: [9fans] apropos of the glendix post

[lucio]

> Is there anyone out there who's gonna be in Volos and wants to
> participate in a driver-writing BOF?

Me, kinda goes without saying.

++L

Re: [9fans] apropos of the glendix post

[Francisco J Ballesteros]

me too.
Not that I wrote any driver since long ago, but I'm interested.

On Fri, Sep 5, 2008 at 6:33 AM,  <lucio@proxima.alt.za> wrote:
>> Is there anyone out there who's gonna be in Volos and wants to
>> participate in a driver-writing BOF?
>
> Me, kinda goes without saying.
>
> ++L
>
>
>

Re: [9fans] apropos of the glendix post

[David Leimbach]

[-- Attachment #1: Type: text/plain, Size: 902 bytes --]

On Fri, Sep 5, 2008 at 1:59 AM, Francisco J Ballesteros <nemo@lsub.org>wrote:

> me too.
> Not that I wrote any driver since long ago, but I'm interested.


I'd be pretty interested in learning how to write device drivers for *any*
OS at this point.  I've take a few passes at drivers for things in the past
in Plan 9, unsuccessfully, then always seemed to get very busy at work and
had to put stuff like that aside.  (I looked at OHCI USB drivers for Plan 9
at one point, but never got very far with it...)

I think a lot of my problem was not understanding the kernel's structures,
and the right way to do things.

Dave


>
>
> On Fri, Sep 5, 2008 at 6:33 AM,  <lucio@proxima.alt.za> wrote:
> >> Is there anyone out there who's gonna be in Volos and wants to
> >> participate in a driver-writing BOF?
> >
> > Me, kinda goes without saying.
> >
> > ++L
> >
> >
> >
>
>

[-- Attachment #2: Type: text/html, Size: 1531 bytes --]

Re: [9fans] apropos of the glendix post

[ron minnich]

On Fri, Sep 5, 2008 at 8:36 AM, Mathieu Lonjaret <lejatorn@gmail.com> wrote:
> Same here.
> I haven't written any serious code in a long time but I'd be very
> interested at least in a tutorial on how to write a file server, or the
> basics steps to write a driver.
>

If you are all planning to do this, you should at least study a few
drivers before you get there. If you go in cold, you're going to waste
an awful lot of time.

ron

Re: [9fans] apropos of the glendix post

[Mathieu Lonjaret]

[-- Attachment #1: Type: text/plain, Size: 197 bytes --]

Same here.
I haven't written any serious code in a long time but I'd be very
interested at least in a tutorial on how to write a file server, or the
basics steps to write a driver.

Mathieu.

[-- Attachment #2: Type: message/rfc822, Size: 3689 bytes --]

From: "Francisco J Ballesteros" <nemo@lsub.org>
To: "Fans of the OS Plan 9 from Bell Labs" <9fans@9fans.net>
Subject: Re: [9fans] apropos of the glendix post
Date: Fri, 5 Sep 2008 10:59:32 +0200
Message-ID: <8ccc8ba40809050159p6258b587jfdadce676bd10927@mail.gmail.com>

me too.
Not that I wrote any driver since long ago, but I'm interested.

On Fri, Sep 5, 2008 at 6:33 AM,  <lucio@proxima.alt.za> wrote:
>> Is there anyone out there who's gonna be in Volos and wants to
>> participate in a driver-writing BOF?
>
> Me, kinda goes without saying.
>
> ++L
>
>
>

Re: [9fans] apropos of the glendix post

[Zak B. Elep]

uriel99@gmail.com (Uriel) writes:

> Sounds like a really cool idea, maybe somebody with more experience in
> the subject wants to give a short introduction, and then people can
> start to work on their favorite wanted driver (alone or in groups)
> while the 'expert'(s) in charge of the session can answer questions
> and help whoever gets stuck.

I've never written a driver for any piece of hardware before, but I'm
interested to learn.  My C is a little rusty, but that's why I've picked
up nemo's book again to reacquaint myself with Plan 9. :)  I hope such a
session pushes through, even if only on IRC.

--
  I like the idea of 256 bits, though: 32 for the (Unicode) character leaves
  room for 224 Bucky bits, which ought to be enough for anyone.
				-- Roland Hutchinson, in alt.folklore.computers

Re: [9fans] apropos of the glendix post

[ron minnich]

if we had common hardware it might be easier. You can get via c800
boards for $99.

Maytbe you can get a bundle and include the cost in the workshop.
Spend a day hacking on drivers. It ought to be a day.

ron