From mboxrd@z Thu Jan 1 00:00:00 1970 From: robherring2@gmail.com (Rob Herring) Date: Fri, 01 Oct 2010 13:47:15 -0500 Subject: [PATCH 3/3] ARM: twd_smp: add clock api support In-Reply-To: <20101001183431.GB22533@n2100.arm.linux.org.uk> References: <1285886952-30888-1-git-send-email-robherring2@gmail.com> <1285886952-30888-3-git-send-email-robherring2@gmail.com> <4CA5417B.5020200@gmail.com> <4CA551FF.3080100@gmail.com> <4CA614AA.8090305@gmail.com> <20101001183431.GB22533@n2100.arm.linux.org.uk> Message-ID: <4CA62CB3.1060909@gmail.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 10/01/2010 01:34 PM, Russell King - ARM Linux wrote: > On Fri, Oct 01, 2010 at 11:22:31AM -0700, Colin Cross wrote: >> The timer APIs require that a timer take at least as long as requested >> to fire. If you change the TWD prescaler before the CPU frequency >> increases, or after the CPU frequency decreases, the timer interrupt >> will not fire until a small time after the requested time. > > That maybe - and for TWD which is used for local timer purposes only, > that's fine. What wouldn't be fine is if your clocksource counter > changes frequency... > > In any case, I'm not sure that passing in a struct clk pointer is > really the right way to go - the point of the clk API is to remove > such passing of static structures. > > How about instead using clk_get_sys() to obtain a clock for the TWD? Okay. Opinions on the name: periphclk, arm_twd, twd? > > Do we need separate connection IDs for each TWD, or are all TWDs in > a system always clocked together? A9 is 1 clock for all. Probably is likely that they will always be symmetric. Rob