First of all, thanks for your aswer,
Gilles Chanteperdrix a écrit :
>>  I patched the kernel with the arm1 patch (from the ARM website), the
>>  board non-rt patch enabling serial and the macb ethernet controller and
>>  finnally tried to apply the ipipe arm patch and resolving failures.
> At first sight, you call spin_lock_irqsave in __ipipe_mach_set_dec.
> This is illegal, __ipipe_mach_set_dec may (and will) be called from
> real-time context, you should be using the _hw variants. I do not know
> if it is what causes the oops, but it would cause an oops later.
I based my board patch on integrator's and s3c24xx code and they use
spin_lock_irqsave in __ipipe_mach_set_dec, but I found local_irq_save_hw
used on an AT91 patch.
Finally the change did nothing, are the integrator and s3c24xx patchs
up-to-date ?

Should I drop the arm1 patch from ARM ? and start from a fresh vanilla
kernel ?

> I am not sure it makes sense to set the timer auto-reload mode when
> the timer is used in one-shot mode: imagine that the timer is
> programmed in one-shot mode, it will tick several time before the
> interrupt handler is called.
Integrator seems using the timer in one-shot mode and changes the
prescaler with the delay value. But won't I loose to much timer ticks
between two reloads ?

Neil Armstrong
Neotion Sophia Antipolis
Stagiaire Mars-Septembre 2008 Polytech'Nice-Sophia
fn:Neil Armstrong
org:Neotion;Neotion Sophia Antipolis
email;internet:[EMAIL PROTECTED]
title;quoted-printable:Ing=C3=A9nieur Software Embarqu=C3=A9
note:PGP 0x1166F485

Attachment: signature.asc
Description: OpenPGP digital signature

Xenomai-core mailing list

Reply via email to