Hey Andrew, On Thu, Sep 26, 2013 at 10:24:04AM +0200, Andrew Lunn wrote: > > + /* > + * Wait for clocks to settle down, and then clear all the > + * ratios request and the reload request. > + */ > + udelay(1000); > + reg &= ~(CORE_CLOCK_DIVIDER_RATIO_MASK | CORE_CLOCK_DIVIDER_RATIO_RELOAD); > + writel(reg, corediv->reg); > + udelay(1000); > > > Documentation/timers/timers-howto.txt says: > > SLEEPING FOR ~USECS OR SMALL MSECS ( 10us - 20ms): > * Use usleep_range > Given the set_rate should be protected by spinlocks (as you pointed out), I guess we can't afford using usleep_range (the context is now non-atomic). Now, according to arch/arm/include/asm/delay.h this is the suggestion for udelay and friends: * Use only for very small delays ( < 2 msec). Should probably use a * lookup table, really, as the multiplications take much too long with * short delays. This is a "reasonable" implementation, though (and * the first constant multiplications gets optimized away if the delay is * a constant) Which means we could just use udelay(1000) without much trouble. On the other side, since we know the rate the clock is running, we could use __delay() just as drivers/clk/sunxi/clk-factors.c is doing. __delay() has a much simpler implementation, which feels correct in this case of "wait until the clock settles". Therefore, I'm currently going to try that option. -- Ezequiel García, Free Electrons Embedded Linux, Kernel and Android Engineering http://free-electrons.com -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html