Re: sparc-leon patches round 5 (to follow) (right numbering)

Re: sparc-leon patches round 5 (to follow) (right numbering)

[David Miller]

From: Konrad Eisele <konrad@gaisler.com>
Date: Tue, 23 Jun 2009 11:11:42 +0200

> round 5 had wrong numbering, I'll repost.

I'm applying all of this to sparc-next-2.6 and will push it
out to kernel.org after some build testing.

Thanks!

Re: sparc-leon patches round 5 (to follow) (right numbering)

[David Miller]

From: David Miller <davem@davemloft.net>
Date: Sun, 02 Aug 2009 19:51:40 -0700 (PDT)

> From: Konrad Eisele <konrad@gaisler.com>
> Date: Tue, 23 Jun 2009 11:11:42 +0200
> 
>> round 5 had wrong numbering, I'll repost.
> 
> I'm applying all of this to sparc-next-2.6 and will push it
> out to kernel.org after some build testing.

Ok, I had to revert all of this.

Are you even build testing what you're posting?

arch/sparc/kernel/built-in.o: In function `leon_node_init':
(.init.text+0x29c4): undefined reference to `_amba_init'

I can guess what happened, while respinning these patches you forgot
to do some "git add" commands and thus the file that contains the
amba_init() code is in your tree but not in any of the commits
and thus not in your patches.

When you fix this up and repost the series, do me another favor, try
to build the kernel successfully after each and every patch.  With
LEON enabled, the tree does not build after patch #3 because srmmu.c
doesn't include asm/leon.h yet and thus srmmu_swprobe() isn't declared.

Please walk through this patch set carefully, checking the build for
both LEON and non-LEON every step of the way, and after every commit
run "git status" to make sure there are not stray files not checked
into your tree.

Thanks.

Re: sparc-leon patches round 5 (to follow) (right numbering)

[konrad]

Hi,
Ok, I'll rework the patches. I'm on holiday right now,
but I'll do it after 17.August when I'll come back
to the office. I'll test it for Leon and non-Leon, I had it
working non-Leon also but apparently it broke someway
when working to cleanup the patches.  The
_amba_init() undefined reference is the only reference
outside the arch/sparc directory. I'll protect it with a #ifdef LEON
kind of macro. The thing is that it is
the root for the Leon specific drivers that are currently
in drivers/ambapp, a new directory which I have to figure
out how to get into the mainline kernel. So it is kind of a
chicken and egg problem, without the arch/sparc part
drivers/ambapp (that contains amba_init() and uart and ethernet
drivers) will not be permitted, without  drivers/ambapp arch/sparc
will not compile. Adding all together to the kernel at once will be too
big. So  I thought that first get the Leon part of arch/sparc in
and then figure out how to proceed with drivers/ambapp. 
amba_init will scan the AMBA bus and detect the connected
devices. It's quite a bit of code and not really connected to arch/sparc. 

-- Greetings Konrad

Quoting David Miller <davem@davemloft.net>:
> From: David Miller <davem@davemloft.net>
> Date: Sun, 02 Aug 2009 19:51:40 -0700 (PDT)
>
> > From: Konrad Eisele <konrad@gaisler.com>
> > Date: Tue, 23 Jun 2009 11:11:42 +0200
> >
> >> round 5 had wrong numbering, I'll repost. 
> >
> > I'm applying all of this to sparc-next-2.6 and will push it
> > out to kernel.org after some build testing. 
>
> Ok, I had to revert all of this. 
>
> Are you even build testing what you're posting?
>
> arch/sparc/kernel/built-in.o: In function `leon_node_init':
> (.init.text+0x29c4): undefined reference to `_amba_init'
>
> I can guess what happened, while respinning these patches you forgot
> to do some "git add" commands and thus the file that contains the
> amba_init() code is in your tree but not in any of the commits
> and thus not in your patches. 
>
> When you fix this up and repost the series, do me another favor, try
> to build the kernel successfully after each and every patch.  With
> LEON enabled, the tree does not build after patch #3 because srmmu.c
> doesn't include asm/leon.h yet and thus srmmu_swprobe() isn't declared. 
>
> Please walk through this patch set carefully, checking the build for
> both LEON and non-LEON every step of the way, and after every commit
> run "git status" to make sure there are not stray files not checked
> into your tree. 
>
> Thanks. 
> --
> To unsubscribe from this list: send the line "unsubscribe sparclinux" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: sparc-leon patches round 5 (to follow) (right numbering)

[Thomas Bogendoerfer]

On Mon, Aug 03, 2009 at 03:05:32PM -0400, konrad@gaisler.com wrote:
> big. So  I thought that first get the Leon part of arch/sparc in
> and then figure out how to proceed with drivers/ambapp. amba_init will 
> scan the AMBA bus and detect the connected

how about using one of various XXX_initcall() things from include/linux/init.h
subsys_initcall sounds about right... That way you don't have to mess in the
 sparc code, but could put it into your driver subdir...

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessary a
good idea.                                                [ RFC1925, 2.3 ]

Re: sparc-leon patches round 5 (to follow) (right numbering)

[David Miller]

From: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Date: Mon, 3 Aug 2009 22:05:00 +0200

> On Mon, Aug 03, 2009 at 03:05:32PM -0400, konrad@gaisler.com wrote:
>> big. So  I thought that first get the Leon part of arch/sparc in
>> and then figure out how to proceed with drivers/ambapp. amba_init will 
>> scan the AMBA bus and detect the connected
> 
> how about using one of various XXX_initcall() things from include/linux/init.h
> subsys_initcall sounds about right... That way you don't have to mess in the
>  sparc code, but could put it into your driver subdir...

Right and this is what other bus layers do too.

Re: sparc-leon patches round 5 (to follow) (right numbering)

[Konrad Eisele]

Hi,

 > arch/sparc/kernel/built-in.o: In function `leon_node_init':
 > (.init.text+0x29c4): undefined reference to `_amba_init'

I posted the patchset version 6 (see mailing list). This time
_amba_init() is not called directly. I had to use another aproach
than the XXX_initcall() aproach that Thomas Boegendoerfer suggested:
leon_node_init() is called too early in the boot process
(main()->setup_arch()->paging_init()->prom_build_devicetree()..) and
as far as I understand it there is no xxx_initcall() that gets
called that early => therefore I used a callback instead:

...
void __initdata (*prom_amba_init)(struct device_node *dp, struct device_node ***nextp) = 0;
...
	if (prom_amba_init &&
...
		prom_amba_init(dp, nextp);
	}
...


The callback will be initialized by the _bootloader_.

 > When you fix this up and repost the series, do me another favor, try
 > to build the kernel successfully after each and every patch.  With
 > LEON enabled, the tree does not build after patch #3 because srmmu.c
 > doesn't include asm/leon.h yet and thus srmmu_swprobe() isn't declared.

I didnt really undestand this one: The patch is split up in 5 subpatches
that should logically divide the big patch. There guess are dependencies
among the subpatches. The only way I can think of is to send all 5 subpatches
as one big patch.

Is it ok with the version 6 patches that I just sent on the mailing list
anyway? (that is 5 subpatches)?

-- Greetings Konrad

Re: sparc-leon patches round 5 (to follow) (right numbering)

[David Miller]

From: Konrad Eisele <konrad@gaisler.com>
Date: Mon, 17 Aug 2009 12:27:32 +0200

> Hi,
> 
>> arch/sparc/kernel/built-in.o: In function `leon_node_init':
>> (.init.text+0x29c4): undefined reference to `_amba_init'
> 
> I posted the patchset version 6 (see mailing list). This time
> _amba_init() is not called directly. I had to use another aproach
> than the XXX_initcall() aproach that Thomas Boegendoerfer suggested:
> leon_node_init() is called too early in the boot process
> (main()->setup_arch()->paging_init()->prom_build_devicetree()..) and
> as far as I understand it there is no xxx_initcall() that gets
> called that early => therefore I used a callback instead:

Ok.

I applied all of the Leon patches, thanks a lot.