On 05/15/2014 09:17 PM, Eric Blake wrote:
On 05/15/2014 03:39 AM, Dongsheng Yang wrote:
> As shown in 'man virsh' about schedinfo:
> Note: The cpu_shares parameter has a valid
value range of 0-262144;
This note documents historical kernel limits; if the kernel has changed,
this may be out of date.
...
> +++ b/src/qemu/qemu_driver.c
> @@ -112,6 +112,8 @@ VIR_LOG_INIT("qemu.qemu_driver");
> #define QEMU_SCHED_MAX_PERIOD 1000000LL
> #define QEMU_SCHED_MIN_QUOTA 1000LL
> #define QEMU_SCHED_MAX_QUOTA 18446744073709551LL
> +#define QEMU_SCHED_MIN_SHARES 2LL
> +#define QEMU_SCHED_MAX_SHARES 262144LL
I'm a bit reluctant to add these numbers - if the kernel ever changes
its range again (which HAS happened for some cgroup tunables), then we
are needlessly preventing use of the newer range.
Yes, I hate these numbers too. But the range is defined in
kernel/sched/sched.h
#define MIN_SHARES (1UL << 1)
#define MAX_SHARES (1UL << 18)
and used in scheduler.
shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES));
So we can not access it out from kernel.
As I found there was some numbers for period and quota here, I added
the shares number here.