Re: [PATCH] arm64: kernel: initialize broadcast hrtimer based clock event device

[Will Deacon]

On Thu, May 29, 2014 at 10:23:01AM +0100, Lorenzo Pieralisi wrote:
> On platforms implementing CPU power management, the CPUidle subsystem
> can allow CPUs to enter idle states where local timers logic is lost on power
> down. To keep the software timers functional the kernel relies on an
> always-on broadcast timer to be present in the platform to relay the
> interrupt signalling the timer expiries.
> 
> For platforms implementing CPU core gating that do not implement an always-on
> HW timer or implement it in a broken way, this patch adds code to initialize
> the kernel software broadcast hrtimer upon boot. It relies on a dynamically
> chosen CPU to be always powered-up. This CPU then relays the timer interrupt
> to CPUs in deep-idle states through its HW local timer device.
> 
> On systems with power management capabilities but no functional HW broadcast
> tick device, the hrtimer based clock event device allows the kernel to
> enter high-resolution timer mode, which improves system latencies and saves
> dynamic power.
> 
> The side effect of having a CPU always-on has implications on power management
> platform capabilities and makes CPUidle suboptimal, since at least a CPU is
> kept always in a shallow idle state by the kernel to relay timer interrupts,
> but at least leaves the kernel with a functional system with some working power
> management capabilities.
> 
> The hrtimer based clock event device has lowest possible rating so that,
> if a platform contains a functional HW clock event device with broadcast
> capabilities, that device is always chosen as a tick broadcast device instead
> of the software based one, now present by default.
> 
> Cc: Preeti U Murthy <preeti@xxxxxxxxxxxxxxxxxx>
> Cc: Will Deacon <will.deacon@xxxxxxx>
> Acked-by: Mark Rutland <mark.rutland@xxxxxxx>
> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@xxxxxxx>

Thanks Lorenzo,

  Acked-by: Will Deacon <will.deacon@xxxxxxx>

With this patch applied, cyclictest starts reporting sane numbers and I
no longer see latency-related failures in LTP when it does things like
test timeouts and nanosleep()s.

Will
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/