headers
Rob Landley | 1 Oct 2007 05:23

[PATCH] Add missing entries to Documentation/arm/00-INDEX

From: Rob Landley <rob <at> landley.net>

Add several missing entries to Documentation/arm/00-INDEX

Signed-off-by: Rob Landley <rob <at> landley.net>
---

 Documentation/arm/00-INDEX |   22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff -r f4ca99897c12 Documentation/arm/00-INDEX
--- a/Documentation/arm/00-INDEX	Sat Sep 29 10:00:15 2007 -0700
+++ b/Documentation/arm/00-INDEX	Sun Sep 30 22:16:06 2007 -0500
 <at>  <at>  -4,19 +4,29  <at>  <at>  Booting
 	- requirements for booting
 Interrupts
 	- ARM Interrupt subsystem documentation
+IXP2000
+	- Release Notes for Linux on Intel's IXP2000 Network Processor
 Netwinder
 	- Netwinder specific documentation
+Porting
+       - Symbol definitions for porting Linux to a new ARM machine.
+Setup
+       - Kernel initialization parameters on ARM Linux
 README
 	- General ARM documentation
-SA1100
+SA1100/
 	- SA1100 documentation
(Continue reading)

Permalink | Reply | 0 comments |
headers
Matthieu CASTET | 1 Oct 2007 10:33
Favicon

Re: [RFC, PATCH] Console on ICEDCC

Uwe Kleine-König wrote:
> Hello,
> 
> my company's devkit shipps with a similar driver on top of the serial
> framework.  But actually I think this approach would be better.
> 
I prefer the serial approch :
- you can use the dcc link for other purposes. For example running a 
gdbserver on it. You could even try to run network on it with ppp or slip.

Matthieu

PS : the serial dcc driver from Hyok S. Choi have some issues. For 
example it is very slow in rx in no interrupt mode (we send a char and 
wait for the timer to expire, we should try to do a small busyloop (like 
we do in tx) to know if are still characters to read).

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 0 comments |
headers
Michel Benoit | 1 Oct 2007 13:58
Picon
Favicon

idle, watchdog and AT91SAM9260

Hello list,

What happens when the kernel has nothing to do and the idle task is
running.  Does the kernel make a call into arm or platform code in
order to set the CPU in idle/standby state?
If so can someone please point me to the location in the code where
this happens.

CPU idle/standby state which blocks until the next interrupt is often
used to save power.

I am running an arm kernel on a AT91SAM9260 cpu.  The watchdog timer
in this SoC has a bit that controls whether or not the watchdog
increments while the cpu is idle state.

Unfortunately this bit can only be written once which means:

-If the watchdog gets disabled while the idle task is running then the
watchdog is only useful when the processor is busy.

-If the watchdog is enabled during idle/standby mode then its not
possible to put the CPU is idle/standby state when the device should
go down into a low power sleep.

Any tips on how to solve this problem?

Michel

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
(Continue reading)

Permalink | Reply | 13 comments |
headers
Russell King - ARM Linux | 1 Oct 2007 14:54
Picon

Re: idle, watchdog and AT91SAM9260

On Mon, Oct 01, 2007 at 01:58:04PM +0200, Michel Benoit wrote:
> What happens when the kernel has nothing to do and the idle task is
> running.  Does the kernel make a call into arm or platform code in
> order to set the CPU in idle/standby state?

Yes.

> If so can someone please point me to the location in the code where
> this happens.

It's CPU dependent.  grep do_idle arch/arm/mm/proc-*.S

> I am running an arm kernel on a AT91SAM9260 cpu.  The watchdog timer
> in this SoC has a bit that controls whether or not the watchdog
> increments while the cpu is idle state.
> 
> Unfortunately this bit can only be written once which means:
> 
> -If the watchdog gets disabled while the idle task is running then the
> watchdog is only useful when the processor is busy.
> 
> -If the watchdog is enabled during idle/standby mode then its not
> possible to put the CPU is idle/standby state when the device should
> go down into a low power sleep.

If you were using the pre-clockevent/clocksource code, I'd say that
the thread will be woken every 10ms, so stroking the watchdog shouldn't
be a problem.

With the clockevent/clocksource code, provided a timer is set to expire
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andrew Victor | 1 Oct 2007 18:04

Re: idle, watchdog and AT91SAM9260

hi Michel,

> What happens when the kernel has nothing to do and the idle task is
> running.  Does the kernel make a call into arm or platform code in
> order to set the CPU in idle/standby state?
> If so can someone please point me to the location in the code where
> this happens.

arch_idle() in include/asm-arm/arch-at91/system.h

   /*
    * Set the processor (CP15) into 'Wait for Interrupt' mode.
    * Unlike disabling the processor clock via the PMC (above)
    *  this allows the processor to be woken via JTAG.
    */

Regards,
  Andrew Victor

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 11 comments |
headers
Michel Benoit | 1 Oct 2007 18:23
Picon
Favicon

Re: idle, watchdog and AT91SAM9260

>
> > What happens when the kernel has nothing to do and the idle task is
> > running.  Does the kernel make a call into arm or platform code in
> > order to set the CPU in idle/standby state?
> > If so can someone please point me to the location in the code where
> > this happens.
>
> arch_idle() in include/asm-arm/arch-at91/system.h
>
>    /*
>     * Set the processor (CP15) into 'Wait for Interrupt' mode.
>     * Unlike disabling the processor clock via the PMC (above)
>     *  this allows the processor to be woken via JTAG.
>     */
>

The result of which is that if the WDIDLEHLT bit is set in the write
once watchdog register WDT_MR of the AT91SAM9260 the clock to the
watchdog timer is disabled while the processor is idle.  Practically
speaking this means that the watchdog timer only increments while the
cpu is working and therefore it can take a long time for it to expire
when the cpu is not busy (even though it is not being kicked).

Unfortunately I would like to not have the watchdog timer active while
I am in low power sleep mode which is the same idle state as enabled
by arch_idle().

As far as I can see I have two options:

1. disable the arch_idle() functionality = higher power consumption in
(Continue reading)

Permalink | Reply | 10 comments |
headers
Russell King - ARM Linux | 1 Oct 2007 19:01
Picon

Re: idle, watchdog and AT91SAM9260

On Mon, Oct 01, 2007 at 06:23:53PM +0200, Michel Benoit wrote:
> 2. not set WDIDLEHLT and wake up from low power sleep mode every < 16
> seconds to kick the watchdog = higher power consumption in low power
> sleep mode.

This is a silly concern - you're going to be woken up far more frequently
than once very 16 seconds.  It's currently about once every second for
the kernel to perform various internal maintainence tasks.

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 0 comments |
headers
Marc Singer | 1 Oct 2007 17:47
Favicon

Re: Platform device driver sequence...

Gianluca wrote:
> Hello list,
> Maybe this is not strictly related to ARM linux ML, but as soon as I am using 
> a linux ARM machine, I hope to find somebody can point/help me out to proceed 
> further.
>
> Here is my concern: our board has an I2C memory device accessed via /dev/nvram 
> interface, and it stores the FrameBuffer timings/resolution of 4 LCD display 
> can be connected (one at time) to this board.
>
> I would like _NOT_ enable the framebuffer driver during early bootup just to 
> not break any LCD Electronics due to a bad timings/frequency, but only 
> _AFTER_ probing the i2c device and activate the framebuffer with the correct
> value found in the eeprom memory.
>
> Ofcourse I have to program the eeprom memory with the correct LCD timing 
> *BEFORE* launching the kernel...
> ;-)
>
> Can the fb driver delayed after another driver? Where I have to check this 
> out? There is a SURE sequence?
>
> Thank you for any help/clue,
>
> Best regards,
>   
You may be able to load the FB driver as a module to delay startup until 
after you I2C has probed.

However, I think that Russell is on the better track.
(Continue reading)

Permalink | Reply | 2 comments |
headers
Dmitry Shmidt | 2 Oct 2007 01:29
Picon
Favicon

Two SD/SDIO devices share one SD controller

Hi,

Our OMAP system has one SD/SDIO controller, multiplexer and SDCard and
SDIO device connected through this multiplexer to controller.
Also there is a GPIO line connected to multiplexer to choose between
the SDCard or SDIO device.
Where is the proper place in new SD/SDIO stack to assert this line ?

Thanks,

Dmitry

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 0 comments |
headers
Marc Pignat | 2 Oct 2007 07:42
Picon
Favicon

Re: idle, watchdog and AT91SAM9260

Hello!

On Monday 01 October 2007, Michel Benoit wrote:
> >
...
> 1. disable the arch_idle() functionality = higher power consumption in
> normal running mode
> 
> 2. not set WDIDLEHLT and wake up from low power sleep mode every < 16
> seconds to kick the watchdog = higher power consumption in low power
> sleep mode.
> 
> Are there any other options?
I think you're mixing 2 things, idle mode and sleep mode.

I agree with Russel, the watchdog should run while you're in idle mode and you
will never get 16 seconds in idle.
When you're in sleep mode (echo standby > /sys/power/state), the watchdog
should be simply disable (see your watchdog suspend/resume functions).

> 
> I need to make some measurements to see how much power consumption is
> affected but I am leaning towards option 2 because I will get the
> added security of an active watchdog while the device is sleeping.  If
> something screws up during sleep or wake-up the device will reset.
Your device will consume more, for no really more security.

Or perhaps I'm missing something, what do you call "low power sleep" ?

Regards
(Continue reading)

Permalink | Reply | 1 comment |

Gmane