[9fans] 9vx as a perfect proto environment

[9fans] 9vx as a perfect proto environment

[ron minnich]

I think 9vx is now the equivalent of software tools from the old days.

It's fast. But the big beauty of it for me is that in vx32/src/9vx/a
is pretty much a plan 9 kernel in plan 9 C vernacular. I just spent an
easy short time prototyping some new stuff that I can now drop into a
real plan 9 kernel for Blue Gene, no changes needed. The
edit/build/test boot cycle is measured in seconds. The fact that we
have a friendly path via codereview(1) and bitbucket is the icing on
the cake.

I know some folks who are using 9vx to get the hang of plan 9. Their
9vx window is always up. It's a great way to ease into the system.

So we have some nice paths. For many unix-like systems, we have
plan9ports. For more of the "real deal", we have 9vx -- and you can
proto some of your kernel code in that environment. And we've got some
very nice ARM platforms coming along, such as the beagleboard, which
are good enough to be a desktop. USB is in, bluetooth is coming ...
things are looking good!

ron

Re: [9fans] 9vx as a perfect proto environment

[Lyndon Nerenberg - VE6BBM/VE7TFX]

> such as the beagleboard, which
> are good enough to be a desktop

Ethernet? My kingdom for Ethernet on one of those!

Is USB Ethernet really viable? It would be nice to hear from anyone
actually doing it (with performance numbers).

--lyndon

Re: [9fans] 9vx as a perfect proto environment

[Richard Miller]

I agree - 9vx is great for experimentation, especially thanks to its
effortless access to the underlying host file system.

I've had less success using it for "real work", at least on MacOS
10.5 and 10.6 - for example running a venti server eventually
results in something like
  9vx panic: sigsegv on cpu7

[No, I don't have an 8-core macbook ...]

Re: [9fans] 9vx as a perfect proto environment

[blstuart]

> I've had less success using it for "real work", at least on MacOS
> 10.5 and 10.6 - for example running a venti server eventually
> results in something like
>   9vx panic: sigsegv on cpu7

I've seen similar on FreeBSD, though I don't think I've see it
on Linux.  I use 9vx pretty much all the time as my terminal
and stand-alone when I'm not connected to my network.  In
those uses, a crash once or twice a day hasn't been bad enough
to make me investigate.

BLS

Re: [9fans] 9vx as a perfect proto environment

[Mathieu L.]

[-- Attachment #1: Type: text/plain, Size: 495 bytes --]

I've definitely seen it on linux.
I'm using 9vx as my primary development platform, and even though it's
fantastic to have a Plan 9 environment so easily setup and integrated
to the host OS, it has its limits. For example it will crash if your
badly written program tries to read on a hanged up connection, or it
will freeze if you run a buggy program in acid with truss.

Still, it's so convenient compared to dual booting or qemu that I prefer
sticking with it.

Cheers,
Mathieu


[-- Attachment #2: Type: message/rfc822, Size: 4004 bytes --]

From: blstuart@bellsouth.net
To: 9fans@9fans.net
Subject: Re: [9fans] 9vx as a perfect proto environment
Date: Sat, 26 Sep 2009 09:52:23 -0500
Message-ID: <46da41288d9f1b334030ffde38b09d9b@bellsouth.net>

> I've had less success using it for "real work", at least on MacOS
> 10.5 and 10.6 - for example running a venti server eventually
> results in something like
>   9vx panic: sigsegv on cpu7

I've seen similar on FreeBSD, though I don't think I've see it
on Linux.  I use 9vx pretty much all the time as my terminal
and stand-alone when I'm not connected to my network.  In
those uses, a crash once or twice a day hasn't been bad enough
to make me investigate.

BLS

Re: [9fans] 9vx as a perfect proto environment

[Tim Newsham]

> I'm using 9vx as my primary development platform, and even though it's
> fantastic to have a Plan 9 environment so easily setup and integrated
> to the host OS, it has its limits. For example it will crash if your
> badly written program tries to read on a hanged up connection, or it
> will freeze if you run a buggy program in acid with truss.

is this with the latest 9vx or the bin snapshot?  I know the
snapshot didnt ignore SIGPIPE which would cause the whole vm to
shutdown when writing to a socket that shut down.  However, that
should be fixed in the latest sources.

> Mathieu

Tim Newsham
http://www.thenewsh.com/~newsham/

Re: [9fans] 9vx as a perfect proto environment

[Russ Cox]

> It's fast. But the big beauty of it for me is that in vx32/src/9vx/a
> is pretty much a plan 9 kernel in plan 9 C vernacular. I just spent an
> easy short time prototyping some new stuff that I can now drop into a
> real plan 9 kernel for Blue Gene, no changes needed. The
> edit/build/test boot cycle is measured in seconds. The fact that we
> have a friendly path via codereview(1) and bitbucket is the icing on
> the cake.

This is definitely one of my favorite things about 9vx.
It's a great environment for doing kernel hacking.

> [comments about instability]

I don't think there's any inherent reason why 9vx must be unstable,
but it certainly has a couple bugs.  I haven't had the time to track
them down and fix them, but I'm always happy to point in the
right direction if you can reproduce one.  There have been a
few reports about it dying with cryptic errors from vx32.  I'd like
to track those down but a reproducible test case is an absolute
requirement for the gritty low-level code at the bottom.

The fact that 9vx works as well as it does has always made me
feel like I was cheating.  It feels like it should be impossible
or at least much harder, and yet there it is, and most things run.

Russ

Re: [9fans] 9vx as a perfect proto environment

[blstuart]

> I don't think there's any inherent reason why 9vx must be unstable,
> but it certainly has a couple bugs.  I haven't had the time to track
> them down and fix them, but I'm always happy to point in the
> right direction if you can reproduce one.  There have been a
> few reports about it dying with cryptic errors from vx32.  I'd like
> to track those down but a reproducible test case is an absolute
> requirement for the gritty low-level code at the bottom.

I wish I could isolate a repeatable case.  For me, the times I see it
are times like coming out a screensaver, or bringing the window up
from under firefox.  But it's not always related to redrawing the
screen.  I also sometimes see it when I do things like click on a
mail messsage to view it.  If there's anything consistent, it seems
to be triggered by X activity rather than CPU activity, but that's
a general impression without supporting data.  The fact that I
usually have upas/fs running means that network activity could
well be connected.

> The fact that 9vx works as well as it does has always made me
> feel like I was cheating.  It feels like it should be impossible
> or at least much harder, and yet there it is, and most things run.

Cheating or not, I've found it to be one of the most interesting
and useful things I've come across in years.  So thanks again for
it.

BLS

Re: [9fans] 9vx as a perfect proto environment

[Iruata Souza]

On Sun, Sep 27, 2009 at 2:14 PM, Russ Cox <rsc@swtch.com> wrote:
>> It's fast. But the big beauty of it for me is that in vx32/src/9vx/a
>> is pretty much a plan 9 kernel in plan 9 C vernacular. I just spent an
>> easy short time prototyping some new stuff that I can now drop into a
>> real plan 9 kernel for Blue Gene, no changes needed. The
>> edit/build/test boot cycle is measured in seconds. The fact that we
>> have a friendly path via codereview(1) and bitbucket is the icing on
>> the cake.
>
> This is definitely one of my favorite things about 9vx.
> It's a great environment for doing kernel hacking.
>
>> [comments about instability]
>
> I don't think there's any inherent reason why 9vx must be unstable,
> but it certainly has a couple bugs.  I haven't had the time to track
> them down and fix them, but I'm always happy to point in the
> right direction if you can reproduce one.  There have been a
> few reports about it dying with cryptic errors from vx32.  I'd like
> to track those down but a reproducible test case is an absolute
> requirement for the gritty low-level code at the bottom.
>
> The fact that 9vx works as well as it does has always made me
> feel like I was cheating.  It feels like it should be impossible
> or at least much harder, and yet there it is, and most things run.
>

the new boot code i am working on was written in 9vx.

i found it - and other programs as well - to hang on one of my
machines, but i still think it occurs because of ubuntu + what appears
to be a bad video driver + pulseaudio + X + ...

if that seems familiar to anyone, let me know that i'm not alone.

iru

Re: [9fans] 9vx as a perfect proto environment

[Ethan Grammatikidis]

On Fri, 25 Sep 2009 18:47:46 -0600
Lyndon Nerenberg - VE6BBM/VE7TFX  <lyndon@orthanc.ca> wrote:

> > such as the beagleboard, which
> > are good enough to be a desktop
>
> Ethernet? My kingdom for Ethernet on one of those!
>
> Is USB Ethernet really viable? It would be nice to hear from anyone
> actually doing it (with performance numbers).
>
> --lyndon
>
>

USB eth works well for me.  I use it between my desktop & PDA, both
Linux machines, and have successfully (if a little slowly) dumped the
PDA's 6GB micro-hard-drive over the connection.  I think I measured
the speed at 4Gb/s once, but that could have been limited by the PDA's
drive.

My biggest problem with it was the start-up delays caused by having a
bloated Linux system on the PDA, :) but now I've removed Gtk+ and all
traces of the GPE "PDA environment" performance is more than
acceptable.

--
Ethan Grammatikidis

Those who are slower at parsing information must
necessarily be faster at problem-solving.

Re: [9fans] 9vx as a perfect proto environment

[erik quanstrom]

> USB eth works well for me.  I use it between my desktop & PDA, both
> Linux machines, and have successfully (if a little slowly) dumped the
> PDA's 6GB micro-hard-drive over the connection.  I think I measured
> the speed at 4Gb/s once, but that could have been limited by the PDA's
> drive.

over a 480mbit connection?  that's a trick.

- erik

Re: [9fans] 9vx as a perfect proto environment

[Mathieu L.]

[-- Attachment #1: Type: text/plain, Size: 188 bytes --]

Indeed, I've just retried with the latest pull from hg and that problem
is gone for me, that's brilliant.
That's one less thing to worry about with 9vx, cheers for that. ☺

Mathieu

[-- Attachment #2: Type: message/rfc822, Size: 4244 bytes --]

From: Tim Newsham <newsham@lava.net>
To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net>
Subject: Re: [9fans] 9vx as a perfect proto environment
Date: Sat, 26 Sep 2009 08:45:48 -1000 (HST)
Message-ID: <Pine.BSI.4.64.0909260844540.23231@malasada.lava.net>

> I'm using 9vx as my primary development platform, and even though it's
> fantastic to have a Plan 9 environment so easily setup and integrated
> to the host OS, it has its limits. For example it will crash if your
> badly written program tries to read on a hanged up connection, or it
> will freeze if you run a buggy program in acid with truss.

is this with the latest 9vx or the bin snapshot?  I know the
snapshot didnt ignore SIGPIPE which would cause the whole vm to
shutdown when writing to a socket that shut down.  However, that
should be fixed in the latest sources.

> Mathieu

Tim Newsham
http://www.thenewsh.com/~newsham/

Re: [9fans] 9vx as a perfect proto environment

[Ethan Grammatikidis]

On Wed, 30 Sep 2009 14:45:18 -0400
erik quanstrom <quanstro@coraid.com> wrote:

> > USB eth works well for me.  I use it between my desktop & PDA, both
> > Linux machines, and have successfully (if a little slowly) dumped the
> > PDA's 6GB micro-hard-drive over the connection.  I think I measured
> > the speed at 4Gb/s once, but that could have been limited by the PDA's
> > drive.
>
> over a 480mbit connection?  that's a trick.
>
> - erik
>

Yeah indeed, LOL.  All I can do on that one is apologise and curse my
memory, I'm sorry.  It's possible I calculated the transfer was
'about' or 'a bit over' 400Mbit/s.  It's not a slow link for the
occasional transfer of a big file or two, nor for VNC or Drawterm at
640x480x16bit.


--
Ethan Grammatikidis

Those who are slower at parsing information must
necessarily be faster at problem-solving.

Re: [9fans] 9vx as a perfect proto environment

[lucio]

> USB eth works well for me.

I have two oldish systems where OHCI works erratically or not at all
with up to date distributions (give or take some skepticism towards
replica/pull).  Are there any suggestions on how to approach
debugging?  I can focus on that over the coming weekend.

++L