RE: MIPS Kernel size

RE: MIPS Kernel size

[Smith, Todd]

Hello Ralf,

I am still interested in some older PDA usage that has limited resources.  I
certainly don't want to hold up or stop current kernel dev but is there a
problem with keeping small kernel and/or userspace limits?

Thank you for the discussion.

Todd Smith <todd.smith@camc.org>  

-----Original Message-----
From: Ralf Baechle
Btw, the -tiny tree of 2.6 has been booted on a 2MB system.  Supposedly
that
was an i386 system so MIPS16 should boot in an even smaller system and a
normal 32-bit MIPS kernel should have enough space to wiggle in 4 megs.

Does anybody on this list actually still care about that small systems?

  Ralf

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 11:12:10AM -0500, Smith, Todd wrote:

> I am still interested in some older PDA usage that has limited resources.  I
> certainly don't want to hold up or stop current kernel dev but is there a
> problem with keeping small kernel and/or userspace limits?

Different tradeoffs.  In general the kernel is optimized for performance,
even at the cost of significant amounts of memory.  As the most infamous
example the kernel is using lots of fairly complex hash and radix trees.

But why would a system that has just a default route need the same kind
of data structures and algorithms it takes to route packets on backbone
router in the default free zone?  Why would you drive a moon rocket to
for shopping?

Linux has generally developped in the direction of larger machines and
higher scalability and sometimes that's causing fairly bad itching.  The
-tiny tree is an attempt to correct this.  It's a development tree but
with the goal of merging changes back into the standard kernel and I
hope much of it will be merged back into 2.6 - 2.8 is too far in the
future to wait for ...

  Ralf

RE: MIPS Kernel size

[Smith, Todd]

Hello Ralf,

I like the idea of the -tiny tree and I hope that these changes get merged
back into the main kernel tree.  One of Linux's historic advantages is that
it can run on older equipment but the trend seems to be supporting only
current equipment

Thanks

Todd Smith

RE: MIPS Kernel size

[Maciej W. Rozycki]

On Mon, 2 Feb 2004, Smith, Todd wrote:

> I like the idea of the -tiny tree and I hope that these changes get merged
> back into the main kernel tree.  One of Linux's historic advantages is that
> it can run on older equipment but the trend seems to be supporting only
> current equipment

 I beg your pardon, but isn't the DECstation old enough?

-- 
+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+--------------------------------------------------------------+
+        e-mail: macro@ds2.pg.gda.pl, PGP key available        +

Re: MIPS Kernel size

[Ralf Baechle]

On Tue, Feb 03, 2004 at 02:41:09PM +0100, Maciej W. Rozycki wrote:

> > I like the idea of the -tiny tree and I hope that these changes get merged
> > back into the main kernel tree.  One of Linux's historic advantages is that
> > it can run on older equipment but the trend seems to be supporting only
> > current equipment
> 
>  I beg your pardon, but isn't the DECstation old enough?

Somehow I'm just sitting in front of two Linux 2.6-supported desktops that
are running pretty well :-)

  Ralf

Re: MIPS Kernel size

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 834 bytes --]

On Mon, 2004-02-02 16:48:43 -0500, Smith, Todd <Todd.Smith@camc.org>
wrote in message <490E0430C3C72046ACF7F18B7CD76A2A56955E@KES.camcare.com>:
> Hello Ralf,
> 
> I like the idea of the -tiny tree and I hope that these changes get merged
> back into the main kernel tree.  One of Linux's historic advantages is that
> it can run on older equipment but the trend seems to be supporting only
> current equipment

It's even running on VAXen. But I admit that I like having more than 8MB
of RAM, though...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: MIPS Kernel size

[Ralf Baechle]

On Tue, Feb 03, 2004 at 03:50:14PM +0100, Jan-Benedict Glaw wrote:

> > I like the idea of the -tiny tree and I hope that these changes get merged
> > back into the main kernel tree.  One of Linux's historic advantages is that
> > it can run on older equipment but the trend seems to be supporting only
> > current equipment
> 
> It's even running on VAXen. But I admit that I like having more than 8MB
> of RAM, though...

At least that explains why you say you don't need heating :-)

  Ralf

Re: MIPS Kernel size

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 1050 bytes --]

On Tue, 2004-02-03 16:18:43 +0100, Ralf Baechle <ralf@linux-mips.org>
wrote in message <20040203151843.GD27848@linux-mips.org>:
> On Tue, Feb 03, 2004 at 03:50:14PM +0100, Jan-Benedict Glaw wrote:
> > > I like the idea of the -tiny tree and I hope that these changes get merged
> > > back into the main kernel tree.  One of Linux's historic advantages is that
> > > it can run on older equipment but the trend seems to be supporting only
> > > current equipment
> > It's even running on VAXen. But I admit that I like having more than 8MB
> > of RAM, though...
> 
> At least that explains why you say you don't need heating :-)

Oh, not really :) It's all desktop sized and "only" sucks off a regular
V.230 line...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: MIPS Kernel size

[Ralf Baechle]

On Tue, Feb 03, 2004 at 05:52:26PM +0100, Jan-Benedict Glaw wrote:

> > At least that explains why you say you don't need heating :-)
> 
> Oh, not really :) It's all desktop sized and "only" sucks off a regular
> V.230 line...

CCITT V.230, "General data communications interface, layer 1" ;-)

  Ralf

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 04:48:43PM -0500, Smith, Todd wrote:

> I like the idea of the -tiny tree and I hope that these changes get merged
> back into the main kernel tree.  One of Linux's historic advantages is that
> it can run on older equipment but the trend seems to be supporting only
> current equipment

Remember Linux development is user driven.  If you think it doesn't run
well enough on your small system then it's time to heat up your
C compiler :-)

  Ralf

MIPS Kernel size

[kip.r2]

What will be the approximate size for a minimal MIPS kernel?

Re: MIPS Kernel size

[Thiemo Seufer]

kip.r2@free.fr wrote:
> What will be the approximate size for a minimal MIPS kernel?

IIRC a 2.4 kernel with the bare minimum to boot on a cobalt
machine is a bit below 600k compressed. You will miss many
drivers on such a System. :-)


Thiemo

Re: MIPS Kernel size

[Ralf Baechle]

On Tue, Jan 27, 2004 at 03:51:31PM +0100, kip.r2@free.fr wrote:

> What will be the approximate size for a minimal MIPS kernel?

Btw, the -tiny tree of 2.6 has been booted on a 2MB system.  Supposedly that
was an i386 system so MIPS16 should boot in an even smaller system and a
normal 32-bit MIPS kernel should have enough space to wiggle in 4 megs.

Does anybody on this list actually still care about that small systems?

  Ralf

Re: MIPS Kernel size

[Karsten Merker]

On Mon, Feb 02, 2004 at 03:09:25PM +0100, Ralf Baechle wrote:
> On Tue, Jan 27, 2004 at 03:51:31PM +0100, kip.r2@free.fr wrote:
> 
> > What will be the approximate size for a minimal MIPS kernel?
> 
> Btw, the -tiny tree of 2.6 has been booted on a 2MB system.  Supposedly that
> was an i386 system so MIPS16 should boot in an even smaller system and a
> normal 32-bit MIPS kernel should have enough space to wiggle in 4 megs.
> 
> Does anybody on this list actually still care about that small systems?

Depends on what you consider "that small". Kernel size is a large
issue for the Cobalt series due to the firmware limits (although
Peter Hortons attempts at a Cobalt bootloader will hopefully help in
this regard). Embedded stuff and PDAs is another field where 2.6
currently seems to pose a problem.

Regards,
Karsten
-- 
#include <standard_disclaimer>
Nach Paragraph 28 Abs. 3 Bundesdatenschutzgesetz widerspreche ich der Nutzung
oder Uebermittlung meiner Daten fuer Werbezwecke oder fuer die Markt- oder
Meinungsforschung.

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 07:43:25PM +0100, Karsten Merker wrote:

> Depends on what you consider "that small". Kernel size is a large
> issue for the Cobalt series due to the firmware limits (although
> Peter Hortons attempts at a Cobalt bootloader will hopefully help in
> this regard). Embedded stuff and PDAs is another field where 2.6
> currently seems to pose a problem.

I really hate that term "embedded" - it's very hard to define.  Anyway,
there's an increasing amount of so-called embedded systems with several
gigabytes of memory and even for much smaller system 2.6 is already
making 2.4 look pale.

The Cobalt case is special; it's firmware could almost be the definition
of the term crap ...

  Ralf

Re: MIPS Kernel size

[Geert Uytterhoeven]

On Mon, 2 Feb 2004, Ralf Baechle wrote:
> On Mon, Feb 02, 2004 at 07:43:25PM +0100, Karsten Merker wrote:
> > Depends on what you consider "that small". Kernel size is a large
> > issue for the Cobalt series due to the firmware limits (although
> > Peter Hortons attempts at a Cobalt bootloader will hopefully help in
> > this regard). Embedded stuff and PDAs is another field where 2.6
> > currently seems to pose a problem.
>
> I really hate that term "embedded" - it's very hard to define.  Anyway,
> there's an increasing amount of so-called embedded systems with several
> gigabytes of memory and even for much smaller system 2.6 is already
> making 2.4 look pale.

Indeed. Today's embedded can easily be larger than desktop/server from only
a few years ago...

Fortunately, system size is still important for some applications. Witness the
existence of a System Size Working Group in the CE Linux Forum. So yes, some
people still care.

> The Cobalt case is special; it's firmware could almost be the definition
> of the term crap ...

Can't you use the firmware to load a second stage booter, which can load larger
kernels?

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 08:38:18PM +0100, Geert Uytterhoeven wrote:

> a few years ago...
> 
> Fortunately, system size is still important for some applications. Witness the
> existence of a System Size Working Group in the CE Linux Forum. So yes, some
> people still care.

All but one of the new MIPS systems I got in the past years support memory
sizes in the GB range ...

> > The Cobalt case is special; it's firmware could almost be the definition
> > of the term crap ...
> 
> Can't you use the firmware to load a second stage booter, which can load
> larger kernels?

That would be possible - but would still leave all the other limitations
of the original firmware such as not supporting netboots.  Peter Horten
seem to have worked on something better.

  Ralf

Re: MIPS Kernel size

[Geert Uytterhoeven]

On Mon, 2 Feb 2004, Ralf Baechle wrote:
> On Mon, Feb 02, 2004 at 08:38:18PM +0100, Geert Uytterhoeven wrote:
> > > The Cobalt case is special; it's firmware could almost be the definition
> > > of the term crap ...
> >
> > Can't you use the firmware to load a second stage booter, which can load
> > larger kernels?
>
> That would be possible - but would still leave all the other limitations
> of the original firmware such as not supporting netboots.  Peter Horten
> seem to have worked on something better.

Unles the second stage booter supports netboots. I don't know anything about
the Cobalt firmware, but just using it to load something like uBoot should be
possible, right?

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 08:50:38PM +0100, Geert Uytterhoeven wrote:

> > That would be possible - but would still leave all the other limitations
> > of the original firmware such as not supporting netboots.  Peter Horten
> > seem to have worked on something better.
> 
> Unles the second stage booter supports netboots. I don't know anything about
> the Cobalt firmware, but just using it to load something like uBoot should be
> possible, right?

Oh, it boots everything as long as it's a compress ELF executable named
/vmlinux.gz on a PC-style partitioned disk with an ext2 filesystem
(must be revision 0!) on it's first partition or it uploaded with cu(1)'s
send file command at 115200bps, 8N1.  No luxury, whistles or bells at
all.

  Ralf

Re: MIPS Kernel size

[Thiemo Seufer]

Ralf Baechle wrote:
> On Mon, Feb 02, 2004 at 08:50:38PM +0100, Geert Uytterhoeven wrote:
> 
> > > That would be possible - but would still leave all the other limitations
> > > of the original firmware such as not supporting netboots.  Peter Horten
> > > seem to have worked on something better.
> > 
> > Unles the second stage booter supports netboots. I don't know anything about
> > the Cobalt firmware, but just using it to load something like uBoot should be
> > possible, right?
> 
> Oh, it boots everything as long as it's a compress ELF executable named
> /vmlinux.gz

AFAIK it also boots such a kernel from a NFS(!) export.

> on a PC-style partitioned disk with an ext2 filesystem
> (must be revision 0!) on it's first partition or it uploaded with cu(1)'s
> send file command at 115200bps, 8N1.  No luxury, whistles or bells at
> all.


Thiemo

Re: MIPS Kernel size

[Ralf Baechle]

On Mon, Feb 02, 2004 at 09:07:42PM +0100, Thiemo Seufer wrote:

> > Oh, it boots everything as long as it's a compress ELF executable named
> > /vmlinux.gz
> 
> AFAIK it also boots such a kernel from a NFS(!) export.

Well possible - the firmware contains a Linux kernel of approx 2.0.30
vintage, so that would have been easy to arrange.  When I kernel work for
Cobalt however booting from disk and over serial were the only two
available options.

  Ralf

Re: MIPS Kernel size

[Peter Horton]

On Mon, Feb 02, 2004 at 08:38:18PM +0100, Geert Uytterhoeven wrote:
> 
> > The Cobalt case is special; it's firmware could almost be the definition
> > of the term crap ...
> 
> Can't you use the firmware to load a second stage booter, which can load larger
> kernels?
> 

Yes, but you still have to maintain a old style EXT2 partition just for
the firmware.

Besides, after a while your second stage loader starts to look like a
boot loader anyway, once you've added network booting of large kernels,
kernel command line support and the other things the original firmware
was missing.

P.