Inte IOP Xscale updates

Inte IOP Xscale updates

[Linus Walleij]

Hi XScale folks,

I haven't really seen any functional patches or confirmation that the
XScale machines in arch/arm/iop* are even booting since 2010 or
so.

Can you confirm that you boot-test these machines regularly?

I am worried that we maintain code that is not used or even working
see.

What are your thought about multiplatform and device tree? I am
pretty certain that if the iop machines were converted to device tree
we could collapse these machines into one directory with few files,
but I may be wrong.

Yours,
Linus Walleij

Inte IOP Xscale updates

[Russell King - ARM Linux]

On Tue, Sep 03, 2013 at 02:01:45PM +0200, Linus Walleij wrote:
> Hi XScale folks,
> 
> I haven't really seen any functional patches or confirmation that the
> XScale machines in arch/arm/iop* are even booting since 2010 or
> so.
> 
> Can you confirm that you boot-test these machines regularly?
> 
> I am worried that we maintain code that is not used or even working
> see.
> 
> What are your thought about multiplatform and device tree? I am
> pretty certain that if the iop machines were converted to device tree
> we could collapse these machines into one directory with few files,
> but I may be wrong.

Just because code doesn't get touched doesn't mean it doesn't work.  In
some cases, that means that the code has reached stability.

The IOP platforms are in regular use.  IOP platforms include such
things as TheCUS machines such as the N2100 and others.  The two N2100s
I have here run 3.1.8 and 3.9.5.  3.9.5 did not require any changes to
arch/arm code for it to work - though it did require a minor change
to the IOP watchdog driver so it would build, namely:

diff --git a/drivers/watchdog/iop_wdt.c b/drivers/watchdog/iop_wdt.c
index d964faf..3841120 100644
--- a/drivers/watchdog/iop_wdt.c
+++ b/drivers/watchdog/iop_wdt.c
@@ -234,7 +234,7 @@ static int __init iop_wdt_init(void)
        /* Configure Watchdog Timeout to cause an Internal Bus (IB) Reset
         * NOTE: An IB Reset will Reset both cores in the IOP342
         */
-       write_wdtsr(IOP13XX_WDTCR_IB_RESET);
+//     write_wdtsr(IOP13XX_WDTCR_IB_RESET);
 
        /* Register after we have the device set up so we cannot race
           with an open */

Inte IOP Xscale updates

[Mikael Pettersson]

Russell King - ARM Linux writes:
 > On Tue, Sep 03, 2013 at 02:01:45PM +0200, Linus Walleij wrote:
 > > Hi XScale folks,
 > > 
 > > I haven't really seen any functional patches or confirmation that the
 > > XScale machines in arch/arm/iop* are even booting since 2010 or
 > > so.
 > > 
 > > Can you confirm that you boot-test these machines regularly?
 > > 
 > > I am worried that we maintain code that is not used or even working
 > > see.
 > > 
 > > What are your thought about multiplatform and device tree? I am
 > > pretty certain that if the iop machines were converted to device tree
 > > we could collapse these machines into one directory with few files,
 > > but I may be wrong.
 > 
 > Just because code doesn't get touched doesn't mean it doesn't work.  In
 > some cases, that means that the code has reached stability.
 > 
 > The IOP platforms are in regular use.  IOP platforms include such
 > things as TheCUS machines such as the N2100 and others.  The two N2100s
 > I have here run 3.1.8 and 3.9.5.

I just booted 3.9, 3.10, and 3.11 on my N2100, without problems.

Compared to the Kirkwood Qnaps the N2100 is quite slow and noisy,
so I haven't actually used mine in years.

As long as they doesn't cause any maintenance problems I think we
should just leave the IOP platforms as-is.

Inte IOP Xscale updates

[Russell King - ARM Linux]

On Sat, Sep 07, 2013 at 12:45:12PM +0200, Mikael Pettersson wrote:
> Compared to the Kirkwood Qnaps the N2100 is quite slow and noisy,
> so I haven't actually used mine in years.

"noisy" is not something I recognise from my two.  The standard fan
was noisy, but that's because its rather cheap.  There's replacements
available on the net - NF-A4x10 - which are dead silent, plug in to
the existing fan connector, and give a proper fan RPM speed so it's
possible to configure the F75375 to do automatic fan speed control on
its own.

Inte IOP Xscale updates

[Mikael Pettersson]

Russell King - ARM Linux writes:
 > On Sat, Sep 07, 2013 at 12:45:12PM +0200, Mikael Pettersson wrote:
 > > Compared to the Kirkwood Qnaps the N2100 is quite slow and noisy,
 > > so I haven't actually used mine in years.
 > 
 > "noisy" is not something I recognise from my two.  The standard fan
 > was noisy, but that's because its rather cheap.  There's replacements
 > available on the net - NF-A4x10 - which are dead silent, plug in to
 > the existing fan connector, and give a proper fan RPM speed so it's
 > possible to configure the F75375 to do automatic fan speed control on
 > its own.

That's good to know, thanks.

Inte IOP Xscale updates

[Linus Walleij]

On Sat, Sep 7, 2013 at 12:45 PM, Mikael Pettersson <mikpe@it.uu.se> wrote:

> As long as they doesn't cause any maintenance problems I think we
> should just leave the IOP platforms as-is.

Well I looked over all machines last week and for IOP there is
one thing that stand out, and it's that iop32x and iop33x are
using <mach/gpio.h> and that is a headache for me as GPIO
maintainer, as I want to consolidate that subsystem and get
away from custom GPIO headers.

But I take it that you're ready to test patches for me, so
I'll hopefully get around to propose something for the
iop32x :-)

Basically I'd move <asm/hardware/iop3xx-gpio.h> down
into drivers/gpio as a gpiochip instead.

Yours,
Linus Walleij

Inte IOP Xscale updates

[Dan Williams]

On Tue, Sep 3, 2013 at 5:01 AM, Linus Walleij <linus.walleij@linaro.org> wrote:
> Hi XScale folks,
>
> I haven't really seen any functional patches or confirmation that the
> XScale machines in arch/arm/iop* are even booting since 2010 or
> so.
>
> Can you confirm that you boot-test these machines regularly?
>

I lost access to my boards when I left, but now that I am back I am
trying to track them down.  iop32x may be the only boards that matter
here, I never saw an iop13xx outside of the development board.

> I am worried that we maintain code that is not used or even working
> see.
>
> What are your thought about multiplatform and device tree? I am
> pretty certain that if the iop machines were converted to device tree
> we could collapse these machines into one directory with few files,
> but I may be wrong.
>

I have not been following device-tree conversions for how they handle
legacy boot loaders?

...I'll just start with reviewing your gpio conversion.

--
Dan

Inte IOP Xscale updates

[Linus Walleij]

On Tue, Sep 10, 2013 at 12:10 AM, Dan Williams <dan.j.williams@intel.com> wrote:

> I have not been following device-tree conversions for how they handle
> legacy boot loaders?

That is handled by way of the appended device tree.

With such platforms I use this:

--enable USE_OF
--enable ARM_APPENDED_DTB
--enable ARM_ATAG_DTB_COMPAT
--enable PROC_DEVICETREE

Then:

cat foo.dtb >> zImage

And it should appear in /proc/devicetree

The resulting zImage can be wrapped into a uImage
if U-boot is used, I use this method too.

> ...I'll just start with reviewing your gpio conversion.

Thanks :-)

Yours,
Linus Walleij

Inte IOP Xscale updates

[Aaro Koskinen]

Hi,

On Tue, Sep 03, 2013 at 02:01:45PM +0200, Linus Walleij wrote:
> I haven't really seen any functional patches or confirmation that the
> XScale machines in arch/arm/iop* are even booting since 2010 or
> so.
> 
> Can you confirm that you boot-test these machines regularly?
> 
> I am worried that we maintain code that is not used or even working
> see.

I have N2100/iop32x in use, and I also test mainline kernels regularly on
it. You don't see many patches probably because it has been very stable.
Not sure where you got the 2010, because from mailing list archives and
even from "git log" you can find booting/testing reports even from 2013.

A.

Inte IOP Xscale updates

[Linus Walleij]

On Tue, Sep 10, 2013 at 2:58 PM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> On Tue, Sep 03, 2013 at 02:01:45PM +0200, Linus Walleij wrote:

>> Can you confirm that you boot-test these machines regularly?
>>
>> I am worried that we maintain code that is not used or even working
>> see.
>
> I have N2100/iop32x in use, and I also test mainline kernels regularly on
> it. You don't see many patches probably because it has been very stable.
> Not sure where you got the 2010, because from mailing list archives and
> even from "git log" you can find booting/testing reports even from 2013.

Thanks, now I know :-)

Sometimes one has to poke around to get the right answers, and
now I know who can test patches for this platform.

Yours,
Linus Walleij