From mboxrd@z Thu Jan 1 00:00:00 1970 From: dick@softplc.com (Dick Hollenbeck) Date: Tue, 04 Jan 2011 00:32:30 -0600 Subject: ks8695_gettimeoffset In-Reply-To: <4D22BDAA.2030902@softplc.com> References: <4D0FA9FB.3060309@softplc.com> <20101220193049.GI28157@n2100.arm.linux.org.uk> <4D18B2B5.5000001@visionsystems.de> <4D22BDAA.2030902@softplc.com> Message-ID: <4D22BEFE.4090001@softplc.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 01/04/2011 12:26 AM, Dick Hollenbeck wrote: > On 01/03/2011 01:55 PM, avictor.za at gmail.com wrote: >> hi, >> >>> Is it possible at all to implement clocksource/clockevents for KS8695? As Dick and "Register Description" already said you cannot read the time register, so clocksource->read cannot be implemented to return ticks elapsed. Or do I see it wrong? >> I'm pretty sure I tested this when I originally submitted the KS8685 >> processor support (May 2007), and the timer registers are readable. >> I would of tested with a userspace program that sat in a tight loop >> calling gettimeofday(). >> >> >> Regards, >> Andrew Victor >> > Thanks Andrew. Can you explain how this is supposed to be a duration, even > if one of these two registers was a readable down counter? > > elapsed = __raw_readl(KS8695_TMR_VA + KS8695_T1TC) + __raw_readl(KS8695_TMR_VA + KS8695_T1PD); > > > Which of the two is a down counter, and why would it be added to one that is not, or is? > > While you are thinking about that, I have a userspace test program that can peek at the two registers using a memory mapped region pointer, and it will tell us if the value is changing over time. Sorry, I should have just ran the program before posting. Both registers are constant at 25000 while running. So my original summary is correct, that this C function is not worth anything. Where do we go from here? Dick