Single zImage for Armadillo and KZM9G

Single zImage for Armadillo and KZM9G

[Laurent Pinchart]

Hello,

I've been trying to boot the same zImage on both Armadillo (r8a7740) and KZM9G 
(sh73a0) and stumbled on CONFIG_MEMORY_START. Armadillo has the memory start 
at 0x40000000 while KZM9G seems to reserver the first 16MB, and thus starts at 
0x41000000.

The memory address and size can be specified per-board in the device tree, but 
a mismatch between the DT memory start and CONFIG_MEMORY_START results in a 
pretty hard boot failure (tested on Armadillo).

Has anyone run into the same issue ? Pointers would be appreciated.

-- 
Regards,

Laurent Pinchart

Re: Single zImage for Armadillo and KZM9G

[Simon Horman]

On Wed, Nov 21, 2012 at 01:38:19AM +0100, Laurent Pinchart wrote:
> Hello,
> 
> I've been trying to boot the same zImage on both Armadillo (r8a7740) and KZM9G 
> (sh73a0) and stumbled on CONFIG_MEMORY_START. Armadillo has the memory start 
> at 0x40000000 while KZM9G seems to reserver the first 16MB, and thus starts at 
> 0x41000000.
> 
> The memory address and size can be specified per-board in the device tree, but 
> a mismatch between the DT memory start and CONFIG_MEMORY_START results in a 
> pretty hard boot failure (tested on Armadillo).
> 
> Has anyone run into the same issue ? Pointers would be appreciated.

The most recent discussion of CONFIG_MEMORY_START that I am aware of was
in a thread that stemmed from "[GIT,PULL] Renesas ARM-based SoC defconfig for
v3.8"

https://patchwork.kernel.org/patch/1627511/

In short, it would be nice to get rid of CONFIG_MEMORY_START to allow,
amongst other things, (valid) values to be read from DT. But there
is some work to be done.

Re: Single zImage for Armadillo and KZM9G

[Laurent Pinchart]

Hi Simon,

On Wednesday 21 November 2012 09:47:14 Simon Horman wrote:
> On Wed, Nov 21, 2012 at 01:38:19AM +0100, Laurent Pinchart wrote:
> > Hello,
> > 
> > I've been trying to boot the same zImage on both Armadillo (r8a7740) and
> > KZM9G (sh73a0) and stumbled on CONFIG_MEMORY_START. Armadillo has the
> > memory start at 0x40000000 while KZM9G seems to reserver the first 16MB,
> > and thus starts at 0x41000000.
> > 
> > The memory address and size can be specified per-board in the device tree,
> > but a mismatch between the DT memory start and CONFIG_MEMORY_START
> > results in a pretty hard boot failure (tested on Armadillo).
> > 
> > Has anyone run into the same issue ? Pointers would be appreciated.
> 
> The most recent discussion of CONFIG_MEMORY_START that I am aware of was
> in a thread that stemmed from "[GIT,PULL] Renesas ARM-based SoC defconfig
> for v3.8"
> 
> https://patchwork.kernel.org/patch/1627511/
> 
> In short, it would be nice to get rid of CONFIG_MEMORY_START to allow,
> amongst other things, (valid) values to be read from DT. But there
> is some work to be done.

Thank you for the information. Is there anyone working (or planning to work in 
the near future) on this ?

-- 
Regards,

Laurent Pinchart

Re: Single zImage for Armadillo and KZM9G

[Simon Horman]

On Wed, Nov 21, 2012 at 02:07:20AM +0100, Laurent Pinchart wrote:
> Hi Simon,
> 
> On Wednesday 21 November 2012 09:47:14 Simon Horman wrote:
> > On Wed, Nov 21, 2012 at 01:38:19AM +0100, Laurent Pinchart wrote:
> > > Hello,
> > > 
> > > I've been trying to boot the same zImage on both Armadillo (r8a7740) and
> > > KZM9G (sh73a0) and stumbled on CONFIG_MEMORY_START. Armadillo has the
> > > memory start at 0x40000000 while KZM9G seems to reserver the first 16MB,
> > > and thus starts at 0x41000000.
> > > 
> > > The memory address and size can be specified per-board in the device tree,
> > > but a mismatch between the DT memory start and CONFIG_MEMORY_START
> > > results in a pretty hard boot failure (tested on Armadillo).
> > > 
> > > Has anyone run into the same issue ? Pointers would be appreciated.
> > 
> > The most recent discussion of CONFIG_MEMORY_START that I am aware of was
> > in a thread that stemmed from "[GIT,PULL] Renesas ARM-based SoC defconfig
> > for v3.8"
> > 
> > https://patchwork.kernel.org/patch/1627511/
> > 
> > In short, it would be nice to get rid of CONFIG_MEMORY_START to allow,
> > amongst other things, (valid) values to be read from DT. But there
> > is some work to be done.
> 
> Thank you for the information. Is there anyone working (or planning to
> work in the near future) on this ?

I was planning to look at it but have not done so yet.
Feel free to jump in if you feel so inclined.

Re: Single zImage for Armadillo and KZM9G

[Laurent Pinchart]

Hi Simon,

On Wednesday 21 November 2012 10:28:34 Simon Horman wrote:
> On Wed, Nov 21, 2012 at 02:07:20AM +0100, Laurent Pinchart wrote:
> > On Wednesday 21 November 2012 09:47:14 Simon Horman wrote:
> > > On Wed, Nov 21, 2012 at 01:38:19AM +0100, Laurent Pinchart wrote:
> > > > Hello,
> > > > 
> > > > I've been trying to boot the same zImage on both Armadillo (r8a7740)
> > > > and KZM9G (sh73a0) and stumbled on CONFIG_MEMORY_START. Armadillo has
> > > > the memory start at 0x40000000 while KZM9G seems to reserver the first
> > > > 16MB, and thus starts at 0x41000000.
> > > > 
> > > > The memory address and size can be specified per-board in the device
> > > > tree, but a mismatch between the DT memory start and
> > > > CONFIG_MEMORY_START results in a pretty hard boot failure (tested on
> > > > Armadillo).
> > > > 
> > > > Has anyone run into the same issue ? Pointers would be appreciated.
> > > 
> > > The most recent discussion of CONFIG_MEMORY_START that I am aware of was
> > > in a thread that stemmed from "[GIT,PULL] Renesas ARM-based SoC
> > > defconfig for v3.8"
> > > 
> > > https://patchwork.kernel.org/patch/1627511/
> > > 
> > > In short, it would be nice to get rid of CONFIG_MEMORY_START to allow,
> > > amongst other things, (valid) values to be read from DT. But there
> > > is some work to be done.
> > 
> > Thank you for the information. Is there anyone working (or planning to
> > work in the near future) on this ?
> 
> I was planning to look at it but have not done so yet.
> Feel free to jump in if you feel so inclined.

I wish I had time to do so :-) If I happen to find an empty space in my 
schedule I'll let you know, but please proceed otherwise.

-- 
Regards,

Laurent Pinchart