Re: TG3(Tigoon) & Kernel 2.4.27

Re: TG3(Tigoon) & Kernel 2.4.27

[Nathanael Nerode]

David S. Miller wrote:
>The tg3 firmware is just a bunch of MIPS instructions.
Well, good to know that.  It's the first I'd heard of it.

>I guess if I ran objdump --disassemble on the image and
>used the output of that in the tg3 driver and "compiled
>that source" they'd be happy.  And this makes the situation
>even more ludicrious.

Before you blithely made this claim, you should have actually tried running
objdump --disassemble on the image.

It's not packed up in a standard ELF/COFF/etc format, so that doesn't
actually work.  You can use simple assembler trickery to pack it up into a
normal object file *if* you have an assembler for mips *and* you know whether
the chip is running little-endian or big-endian (I have no idea). You may
need other information as well.  :-P  Then and only then can you try to
dissassemble it with objdump.  Then you have to separate data in the text
section out from the code in order to get something which actually
approximates source code.

Much simpler for Broadcom to just release the source code.  :-P

-- 
This space intentionally left blank.

Re: TG3(Tigoon) & Kernel 2.4.27

[Jeff Garzik]

Nathanael Nerode wrote:
> David S. Miller wrote:
> 
>>The tg3 firmware is just a bunch of MIPS instructions.
> 
> Well, good to know that.  It's the first I'd heard of it.

> Much simpler for Broadcom to just release the source code.  :-P


The previous generation of devices, driven by the acenic driver, 
actually had a firmware kit that was available to lucky parties. 
Included full firmware source, that you could tweak to your heart's delight.

	Jeff

Re: TG3(Tigoon) & Kernel 2.4.27

[David S. Miller]

On Mon, 30 Aug 2004 18:16:38 -0400
neroden@fastmail.fm (Nathanael Nerode) wrote:

> >I guess if I ran objdump --disassemble on the image and
> >used the output of that in the tg3 driver and "compiled
> >that source" they'd be happy.  And this makes the situation
> >even more ludicrious.
> 
> Before you blithely made this claim, you should have actually tried running
> objdump --disassemble on the image.

I was speaking abstractly.  It is well definted which parts of
the hex arrays are the data, bss, and text sections.  It doesn't
take a rocket scientist to reconstruct things.

Re: TG3(Tigoon) & Kernel 2.4.27

[Andreas Schwab]

neroden@fastmail.fm (Nathanael Nerode) writes:

> You can use simple assembler trickery to pack it up into a
> normal object file *if* you have an assembler for mips *and* you know whether
> the chip is running little-endian or big-endian (I have no idea). You may
> need other information as well.  :-P  Then and only then can you try to
> dissassemble it with objdump.

You don't have to pack it up, objdump can also disassemble raw data (with
--target=binary, --disassemble-all and suitable --architecture).

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

TG3(Tigoon) & Kernel 2.4.27

[Laurent CARON]

Hello

When I try to compile kernel 2.4.27 for one of my servers i get this error.
--------------
drivers/net/net.o(.text+0x17550): In function `tg3_request_firmware': : 
undefined reference to `request_firmware'
drivers/net/net.o(.text+0x17652): In function `tg3_request_firmware': : 
undefined reference to `release_firmware'
-------------

Any clue?

PS: I can include a part of my .config

Re: TG3(Tigoon) & Kernel 2.4.27

[Randy.Dunlap]

On Tue, 24 Aug 2004 17:13:57 +0200 Laurent CARON wrote:

| Hello
| 
| When I try to compile kernel 2.4.27 for one of my servers i get this error.
| --------------
| drivers/net/net.o(.text+0x17550): In function `tg3_request_firmware': : 
| undefined reference to `request_firmware'
| drivers/net/net.o(.text+0x17652): In function `tg3_request_firmware': : 
| undefined reference to `release_firmware'
| -------------
| 
| Any clue?
| 
| PS: I can include a part of my .config

You need to enable CONFIG_EXPERIMENTAL and CONFIG_HOTPLUG
and then CONFIG_FW_LOADER in the Library routines menu.

--
~Randy

Re: TG3(Tigoon) & Kernel 2.4.27

[David S. Miller]

On Tue, 24 Aug 2004 09:25:33 -0700
"Randy.Dunlap" <rddunlap@osdl.org> wrote:

> | --------------
> | drivers/net/net.o(.text+0x17550): In function `tg3_request_firmware': : 
> | undefined reference to `request_firmware'
> | drivers/net/net.o(.text+0x17652): In function `tg3_request_firmware': : 
> | undefined reference to `release_firmware'
> | -------------
> | 
> | Any clue?
> | 
> | PS: I can include a part of my .config
> 
> You need to enable CONFIG_EXPERIMENTAL and CONFIG_HOTPLUG
> and then CONFIG_FW_LOADER in the Library routines menu.

Oh on the contrary, I've never seen a call to request_firmware()
in any copy of the tg3 driver and that's strange being that I'm
the maintainer. :-)

People, if you're going to use patched up kernels, report your
problems to the people you got the kernel from.  Thanks.

Re: TG3(Tigoon) & Kernel 2.4.27

[Joshua Kwan]

David S. Miller wrote:
> Oh on the contrary, I've never seen a call to request_firmware()
> in any copy of the tg3 driver and that's strange being that I'm
> the maintainer. :-)
> 
> People, if you're going to use patched up kernels, report your
> problems to the people you got the kernel from.  Thanks.

As a matter of fact this is most certainly kernel-source-2.4.27 from 
Debian. This problem was recently fixed (as discussed in the mini-thread 
about how CML1 blows chunks.)

For what it's worth to Google, kernel-tree-2.4.27-4 has just hit 
unstable (will reach mirrors by tomorrow) and should have fixed this 
problem.

Sorry for the noise caused to LKML.

-- 
Joshua Kwan

Re: TG3(Tigoon) & Kernel 2.4.27

[Randy.Dunlap]

On Tue, 24 Aug 2004 18:37:41 -0700 Joshua Kwan wrote:

| David S. Miller wrote:
| > Oh on the contrary, I've never seen a call to request_firmware()
| > in any copy of the tg3 driver and that's strange being that I'm
| > the maintainer. :-)
| > 
| > People, if you're going to use patched up kernels, report your
| > problems to the people you got the kernel from.  Thanks.
| 
| As a matter of fact this is most certainly kernel-source-2.4.27 from 
| Debian. This problem was recently fixed (as discussed in the mini-thread 
| about how CML1 blows chunks.)
| 
| For what it's worth to Google, kernel-tree-2.4.27-4 has just hit 
| unstable (will reach mirrors by tomorrow) and should have fixed this 
| problem.
| 
| Sorry for the noise caused to LKML.

Please don't drop people from to: / cc:

--
~Randy

Re: TG3(Tigoon) & Kernel 2.4.27

[Joshua Kwan]

Randy.Dunlap wrote:
> Please don't drop people from to: / cc:

Something is wrong with this MUA. Sorry.

-- 
Joshua Kwan

Re: TG3(Tigoon) & Kernel 2.4.27

[Willy Tarreau]

Hi David,

On Tue, Aug 24, 2004 at 11:34:07AM -0700, David S. Miller wrote:
 
> Oh on the contrary, I've never seen a call to request_firmware()
> in any copy of the tg3 driver and that's strange being that I'm
> the maintainer. :-)
> 
> People, if you're going to use patched up kernels, report your
> problems to the people you got the kernel from.  Thanks.

I think that he's using a debian kernel and thinks he's using a vanilla
kernel while it's not. I've just installed one on an ultra5 and discovered
a directory with about 60 patches, one of which was dedicated to the removal
of this "undesired closed-source tg3 firmware" ! I was happy not having lost
time debugging this on a machine with a tg3 NIC.

I find it a bit odd when distro makers don't clearly put their name in their
kernels, because it confuses people who honnestly think it's a vanilla one.
Indeed, I'm yet not really sure what they put in the kernel that booted my
sparc, but it's not a problem now since I've replaced it ;-)

Cheers,
Willy

Re: TG3(Tigoon) & Kernel 2.4.27

[David S. Miller]

On Wed, 25 Aug 2004 06:45:52 +0200
Willy Tarreau <willy@w.ods.org> wrote:

> I think that he's using a debian kernel

I knew he was using a debian kernel from the moment I saw
the string "tg3_request_firmware()" in the very first email
of this thread.

I pity the poor fool who wishes to netboot his system using
a tg3 card and use an nfsroot with Debian.  Kind of hard to
get the card firmware from the filesystem in that case.

I guess these debian kernel folks replace the BIOS on their
system with one they have the sources for as well.

The tg3 firmware is just a bunch of MIPS instructions.
I guess if I ran objdump --disassemble on the image and
used the output of that in the tg3 driver and "compiled
that source" they'd be happy.  And this makes the situation
even more ludicrious.

Re: TG3(Tigoon) & Kernel 2.4.27

[Joshua Kwan]

David S. Miller wrote:
> I pity the poor fool who wishes to netboot his system using
> a tg3 card and use an nfsroot with Debian.  Kind of hard to
> get the card firmware from the filesystem in that case.

For the record, I made this change to Debian's kernel source, because
the project consensus was that the firmware blobs were not free
according to Debian's Free Software Guidelines.

Now I hope you're also aware that tg3 without the firmware works for
about ~90-95% of all tigon3 based devices.

Anyway, to tell you the truth, I would have preferred not making this
change. But the alternative was impossible: sticking the kernel into our
non-free section. See below for more.

> I guess these debian kernel folks replace the BIOS on their
> system with one they have the sources for as well.

Nope. Good ol' Award is fine enough for me.

Then again, LinuxBIOS would satisfy the need for source, wouldn't it?
;)

> The tg3 firmware is just a bunch of MIPS instructions.
> I guess if I ran objdump --disassemble on the image and
> used the output of that in the tg3 driver and "compiled
> that source" they'd be happy.  And this makes the situation
> even more ludicrious.

I'm curious: run that by debian-legal@lists.debian.org and see what they
say. They're the folks who started all this.

As far as I'm personally concerned, as long as Linus/Marcelo thinks it's
free (i.e. it can be found in BK) and no one but people in Debian have
the boredom/presence of mind/whatever to point out that this breaks our
software guidelines, I find myself siding with the majority who believe
it's free enough. The linux-kernel community seems to be actively
cracking down on blatant GPL violations (see: Linksys WRT54G), and
that's good, but they (perhaps wisely) seem to have overlooked the
supposed firmware-in-file issue, and I figure that does say something.

I'm not clear on the details but what they seem to have deduced is that
as long as the firmware 'variable' is not shipped into Debian as part of
the compiled tg3.o, it's barely Free. See

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=243044

for details of the loophole. To me, it makes no difference...

Interesting to know that it's a "bunch of MIPS instructions."  I'd wager
that eventually, some enterprising, bored hacker would probably come up
with a glorious Free cleanroom implementation of the firmware that you 
need to use a MIPS assembler or C comiler on on at build time.. (AIUI 
there's already some of that stuff in linux-2.6, though they're actually 
the original source code.)

As I said it's all very disgusting (hmm.. déjà vu..) and maybe people 
will stop caring about it someday.

-- 
Joshua Kwan

Re: TG3(Tigoon) & Kernel 2.4.27

[Oliver Neukum]

Am Mittwoch, 25. August 2004 09:47 schrieb Joshua Kwan:
> I'm not clear on the details but what they seem to have deduced is that
> as long as the firmware 'variable' is not shipped into Debian as part of
> the compiled tg3.o, it's barely Free. See
> 
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=243044
> 
> for details of the loophole. To me, it makes no difference...

And it really strains the meaning of "bugreport"

	Regards
		Oliver

Re: TG3(Tigoon) & Kernel 2.4.27

[David S. Miller]

On Wed, 25 Aug 2004 10:34:59 +0200
Oliver Neukum <oliver@neukum.org> wrote:

> Am Mittwoch, 25. August 2004 09:47 schrieb Joshua Kwan:
> > I'm not clear on the details but what they seem to have deduced is that
> > as long as the firmware 'variable' is not shipped into Debian as part of
> > the compiled tg3.o, it's barely Free. See
> > 
> > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=243044
> > 
> > for details of the loophole. To me, it makes no difference...
> 
> And it really strains the meaning of "bugreport"

And I have to assume that you guys did the same thing to the
qlogic scsi drivers too?  That should wipe out about half of
the ia64 boxes out there, ROFL!

Re: TG3(Tigoon) & Kernel 2.4.27

[Ryan Cumming]

[-- Attachment #1: Type: text/plain, Size: 767 bytes --]

On Tuesday 24 August 2004 23:36, you wrote:
> I pity the poor fool who wishes to netboot his system using
> a tg3 card and use an nfsroot with Debian.  Kind of hard to
> get the card firmware from the filesystem in that case.
initramfs?

> The tg3 firmware is just a bunch of MIPS instructions.
> I guess if I ran objdump --disassemble on the image and
> used the output of that in the tg3 driver and "compiled
> that source" they'd be happy.  And this makes the situation
> even more ludicrious.
For GPL compliance, no, that wouldn't work. The GPL states:
"The source code for a work means the preferred form of the work for making 
modifications to it."

Which is likely C or at least assembly with label names and comments in this 
case.

-Ryan

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

Re: TG3(Tigoon) & Kernel 2.4.27

[David S. Miller]

On Mon, 30 Aug 2004 18:14:38 -0700
Ryan Cumming <ryan@spitfire.gotdns.org> wrote:

> On Tuesday 24 August 2004 23:36, you wrote:
> > I pity the poor fool who wishes to netboot his system using
> > a tg3 card and use an nfsroot with Debian.  Kind of hard to
> > get the card firmware from the filesystem in that case.
> initramfs?

Then it's a part of the kernel image.
You can't have it both ways dude.

Re: TG3(Tigoon) & Kernel 2.4.27

[Willy Tarreau]

On Mon, Aug 30, 2004 at 06:14:38PM -0700, Ryan Cumming wrote:
> On Tuesday 24 August 2004 23:36, you wrote:
> > I pity the poor fool who wishes to netboot his system using
> > a tg3 card and use an nfsroot with Debian.  Kind of hard to
> > get the card firmware from the filesystem in that case.
> initramfs?
> 
> > The tg3 firmware is just a bunch of MIPS instructions.
> > I guess if I ran objdump --disassemble on the image and
> > used the output of that in the tg3 driver and "compiled
> > that source" they'd be happy.  And this makes the situation
> > even more ludicrious.
> For GPL compliance, no, that wouldn't work. The GPL states:
> "The source code for a work means the preferred form of the work for making 
> modifications to it."
> 
> Which is likely C or at least assembly with label names and comments in this 
> case.

It would have taken less time for the people who removed the firmware and
will assure support for their users to disassemble this code and put labels
and comments in it. When I was 16, I totally disassembled my PC's bios (8kB)
commented it and labelled it to the point of making it re-assemblable. It
did not take very much time (a week-end), and it was crappy x86 code with
unaligned and mangled code/data + hard-coded stack values for the "calls"
without any RAM usage. Here you have about 2kB of clean MIPS code which is
about 500 instructions (500 lines). I doubt anyone knowing mips assembly
enough would spend more than a week-end putting comments and labels in it.

Now another possibility would be for those zealots to replace the firmware
by an open-source one such as http://alteon.shareable.org/ which BTW contains
all the tools needed for the dirty work.

Anyway, this thread is getting boring...

Regards,
Willy

Re: TG3(Tigoon) & Kernel 2.4.27

[Laurent CARON]

David S. Miller wrote:

>On Tue, 24 Aug 2004 09:25:33 -0700
>"Randy.Dunlap" <rddunlap@osdl.org> wrote:
>
>  
>
>>| --------------
>>| drivers/net/net.o(.text+0x17550): In function `tg3_request_firmware': : 
>>| undefined reference to `request_firmware'
>>| drivers/net/net.o(.text+0x17652): In function `tg3_request_firmware': : 
>>| undefined reference to `release_firmware'
>>| -------------
>>| 
>>| Any clue?
>>| 
>>| PS: I can include a part of my .config
>>
>>You need to enable CONFIG_EXPERIMENTAL and CONFIG_HOTPLUG
>>and then CONFIG_FW_LOADER in the Library routines menu.
>>    
>>
>
>Oh on the contrary, I've never seen a call to request_firmware()
>in any copy of the tg3 driver and that's strange being that I'm
>the maintainer. :-)
>
>People, if you're going to use patched up kernels, report your
>problems to the people you got the kernel from.  Thanks.
>  
>

My mistake, I forgot to mention I was using a debian-source :(

Laurent