[PATCH 0/2] tools: Skip backward time entries


Jiri Olsa
 

hi,
on RHEL7 we are facing timing issues in some bcc-tools
scripts. The time subsystem might return backward timestamp
via bpf_ktime_get_ns and thus screw up latency computation.

This seems to be RHEL7 kernel issue and needs to be fixed,
however meanwhile it'd be nice if affected scripts display
sane values despite the kernel issue.

This patchset harden 2 scripts (xfsslower and ext4dist)
we saw this behaviour so far, to ensure the latency is
always > 0.

thanks,
jirka


---
Jiri Olsa (2):
tools: Skip backward time entries in xfsslower
tools: Skip backward time entries in ext4dist

tools/ext4dist.py | 10 ++++++++--
tools/xfsslower.py | 9 ++++++++-
2 files changed, 16 insertions(+), 3 deletions(-)


Yonghong Song
 

Jiri,

The change looks good. Could you generate a pull request against the
bcc/iovisor github repository?

Thanks!

Yonghong

On Tue, Aug 28, 2018 at 5:01 AM Jiri Olsa <jolsa@...> wrote:

hi,
on RHEL7 we are facing timing issues in some bcc-tools
scripts. The time subsystem might return backward timestamp
via bpf_ktime_get_ns and thus screw up latency computation.

This seems to be RHEL7 kernel issue and needs to be fixed,
however meanwhile it'd be nice if affected scripts display
sane values despite the kernel issue.

This patchset harden 2 scripts (xfsslower and ext4dist)
we saw this behaviour so far, to ensure the latency is
always > 0.

thanks,
jirka


---
Jiri Olsa (2):
tools: Skip backward time entries in xfsslower
tools: Skip backward time entries in ext4dist

tools/ext4dist.py | 10 ++++++++--
tools/xfsslower.py | 9 ++++++++-
2 files changed, 16 insertions(+), 3 deletions(-)



Brendan Gregg
 

I'm ok with the workaround code, but I'd like a comment that clearly explains it for future maintainers and copy-n-pasters. Eg:

// Skip entries with backwards time: temp workaround for #728

Then people know it's a temp workaround, and they can lookup https://github.com/iovisor/bcc/issues/728

Brendan

On Tue, Aug 28, 2018 at 9:05 PM, Y Song <ys114321@...> wrote:
Jiri,

The change looks good. Could you generate a pull request against the
bcc/iovisor github repository?

Thanks!

Yonghong
On Tue, Aug 28, 2018 at 5:01 AM Jiri Olsa <jolsa@...> wrote:
>
> hi,
> on RHEL7 we are facing timing issues in some bcc-tools
> scripts. The time subsystem might return backward timestamp
> via bpf_ktime_get_ns and thus screw up latency computation.
>
> This seems to be RHEL7 kernel issue and needs to be fixed,
> however meanwhile it'd be nice if affected scripts display
> sane values despite the kernel issue.
>
> This patchset harden 2 scripts (xfsslower and ext4dist)
> we saw this behaviour so far, to ensure the latency is
> always > 0.
>
> thanks,
> jirka
>
>
> ---
> Jiri Olsa (2):
>       tools: Skip backward time entries in xfsslower
>       tools: Skip backward time entries in ext4dist
>
>  tools/ext4dist.py  | 10 ++++++++--
>  tools/xfsslower.py |  9 ++++++++-
>  2 files changed, 16 insertions(+), 3 deletions(-)
>
>
>