Re: [PATCH 08/12] xen/time: introduce xen_{save,restore}_steal_clock

[Boris Ostrovsky]

On 5/19/20 7:28 PM, Anchal Agarwal wrote:
> From: Munehisa Kamata <kamatam@xxxxxxxxxx>
>
> Currently, steal time accounting code in scheduler expects steal clock
> callback to provide monotonically increasing value. If the accounting
> code receives a smaller value than previous one, it uses a negative
> value to calculate steal time and results in incorrectly updated idle
> and steal time accounting. This breaks userspace tools which read
> /proc/stat.
>
> top - 08:05:35 up  2:12,  3 users,  load average: 0.00, 0.07, 0.23
> Tasks:  80 total,   1 running,  79 sleeping,   0 stopped,   0 zombie
> Cpu(s):  0.0%us,  0.0%sy,  0.0%ni,30100.0%id,  0.0%wa,  0.0%hi, 0.0%si,-1253874204672.0%st
>
> This can actually happen when a Xen PVHVM guest gets restored from
> hibernation, because such a restored guest is just a fresh domain from
> Xen perspective and the time information in runstate info starts over
> from scratch.
>
> This patch introduces xen_save_steal_clock() which saves current values
> in runstate info into per-cpu variables. Its couterpart,
> xen_restore_steal_clock(), sets offset if it found the current values in
> runstate info are smaller than previous ones. xen_steal_clock() is also
> modified to use the offset to ensure that scheduler only sees
> monotonically increasing number.
>
> Signed-off-by: Munehisa Kamata <kamatam@xxxxxxxxxx>
> Signed-off-by: Anchal Agarwal <anchalag@xxxxxxxxxx>
> ---
>  drivers/xen/time.c    | 29 ++++++++++++++++++++++++++++-
>  include/xen/xen-ops.h |  2 ++
>  2 files changed, 30 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/xen/time.c b/drivers/xen/time.c
> index 0968859c29d0..3560222cc0dd 100644
> --- a/drivers/xen/time.c
> +++ b/drivers/xen/time.c
> @@ -23,6 +23,9 @@ static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate);
>  
>  static DEFINE_PER_CPU(u64[4], old_runstate_time);
>  
> +static DEFINE_PER_CPU(u64, xen_prev_steal_clock);
> +static DEFINE_PER_CPU(u64, xen_steal_clock_offset);


Can you use old_runstate_time here? It is used to solve a similar
problem for pv suspend, isn't it?


-boris