Update to below
The uploaded patch also addresses:
ARM 2.4 kernels don't appear as standard to increment the usec time value
between ticks. Update include/asm-arm/arch-ep93xx/time.h to provide effective
"gettimeoffset". Assign this in extern __inline__ void setup_timer(void) as
gettimeoffset = ep93xx_gettimeoffset;
irqs.h #define NR_IRQS 64 (was 59)
Update to clk restriction on enet.
Ethernet will work at a 12.5MHz clock (CPU and BUS) if the connection is at
10MBit (10baseT) which can be forced by config (haven't done that yet) or by
connection to 10MBit hub (have done that).
In the ep93xx_eth driver the auto-negotiation has been disabled (don't know by
whom) because it is set in Redboot. This is fine until you start changing
the clock or disabling/enabling the phy. So if you intend to change clock
and/or disable the phy then you need to re-enable the auto-negotiate and then
do a ifconfig down, ifconfig up around the clock change/enable phy which
causes the phy to re-establish proper comms. Not yet clear why the clock
change causes the problem as well as the disable/enable, haven't looked at
the Micrel data sheet yet. This is NOT in the patch of 250705.
I suspect there may be other peripherals which may need clk handling if the
main clk is changed but haven't come across any yet. We only use serial and
enet i/o.
Jim Jackson's cctl is very useful for the enable/disable testing
Bob
On Monday 25 July 2005 12:44, Bob Lees wrote:
> All
>
> I have uploaded a patch file to the files area for the linux_ts8 kernel.
> This patch is against the linux_ts8.tar.gz file from the Technologics web
> site.
>
> It has the following changes:
> 1) patched with 2.4.26-lck1 which effectively gives 0/1 scheduler and
> low_latency. Preempt is not available as this is heavily tied into x86
> arch. 0/1 scheduler is always enabled with this patch, low_latency optional
> and can be controlled via sysctl if required.
> 2) back ported ep93xx-rtc.c from 2.6.11.7 diff file. This uses the
> onboard chip rtc for EP93xx. Mainly we used this for load testing using
> 1Hz interrupt and fast clock.
> 3) back ported ep93xx-wdt.c from 2.6.11.7 diff file. This uses the
> onboard chip watchdog. Reads status bits when called so the reason for
> reboot can be determined. Look at code for IOCTL calls to control this
> device.
> 4) added cpufreq driver a la SA11[01]0. The EP93xx can have almost
> continuous frequency change, but we have only implemented fixed frequency
> steps. Mainly due to our requirement to avoid PLL loop changes; this means
> we should avoid PLL loop lockup delays. Main power saving is from reducing
> BUSCLK rather than CPU CLK. Ethernet doesn't work with a BUSCLK less than
> 25MHz, at least 100MBit doesn't. Look at code to see permissable speeds
> for this patch
> Change speed 'echo newspeed > /proc/sys/cpu/0/speed'
>
> This code is running in TS7200 units. It has NOT been tested in a TS7250.
>
> Any feedback gratefully received.
>
> This is the first time I have generated a kernel patch file so usual
> caveats apply, probably great caution should be used:)
>
> Bob
--
V +44 (0) 1296 747667
F +44 (0) 1296 747557
C +44 (0) 7860 406093
Diamond Consulting Services Ltd
Dinton
Aylesbury
Bucks, HP17 8UG
England
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/ts-7000/
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|