Hi Grygorii,
On 12 July 2017 18:32, Grygorii Strashko wrote:
> On 07/12/2017 08:55 AM, Phil Edworthy wrote:
> > Hi,
> >
> > I'm new to rt-linux, so please forgive me if I ask some dumb questions.
> > I have searched for a while, but not found what I am looking for.
> >
> > I'm trying out the rt-linux patches (v4.9.35) on an out-of-tree BSP
> > for a new ARM Cortex A7 device. When I tried cyclictest, it complained
> > that I don't have High Resolution timers, so I checked and sure enough
> > /proc/timer_list showed the timer resolution is 10ms. I've checked my
> > kernel config and I have enabled hrtimers.
> >
> > This device uses the ARM architected timer (CONFIG_ARM_ARCH_TIMER)
> > running at a lowly 6.25MHz.
> >
> > So, how can I tell if a driver is supposed to support hrtimers?
> > If the ARM architected timer driver does support hrtimers, what could
> > stop them being used?
>
>
> Most probably you do not have at least one tick_broadcast timer enabled.
> See:
> tick_check_oneshot_change()->tick_is_oneshot_available()
Many thanks for the pointer, I didn't know where to start looking.
You are right, tick_is_oneshot_available is returning 0 because
in tick_broadcast_oneshot_available () the code determines that the timer doesn't
have CLOCK_EVT_FEAT_ONESHOT. However, the ARM arch timer does enable
that feature.
So does this mean the ARM arch timer is used for something else, thus I cannot
use it for an hrtimer?
> Alternative option for me was:
> disable CONFIG_GENERIC_CLOCKEVENTS_BROADCAST
That's not a menuconfig option, so I assume you mean hack the Kconfig so that
this is not enabled.
Thanks
Phil
��.n��������+%������w��{.n�����{�����ǫ���ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
