Re: [PATCH v2 iproute2-next] utils: add fflush_monitor() helper

[Stephen Hemminger <stephen@networkplumber.org>]

On Thu, 30 Apr 2026 09:35:08 -0600
David Ahern <dsahern@kernel.org> wrote:

> On 4/27/26 2:19 AM, Eric Dumazet wrote:
> > Some fflush() calls only make sense for monitor programs.
> > 
> > For other cases, forcing a flush is expensive.
> > 
> > After this patch, ip, tc and ss are correctly buffering most of their
> > output when redirected to a file.
> > 
> > Signed-off-by: Eric Dumazet <edumazet@google.com>
> > ---
> >  include/utils.h  |  7 +++++++
> >  ip/ipaddress.c   | 12 +++++++-----
> >  ip/iplink.c      |  4 ++--
> >  ip/ipmonitor.c   |  1 +
> >  ip/ipmptcp.c     | 10 +++++-----
> >  ip/ipneigh.c     |  4 ++--
> >  ip/ipnetconf.c   |  2 +-
> >  ip/ipnetns.c     |  2 +-
> >  ip/ipnexthop.c   | 12 ++++++------
> >  ip/iproute.c     |  4 ++--
> >  ip/iprule.c      |  2 +-
> >  ip/iptoken.c     |  2 +-
> >  ip/tcp_metrics.c |  2 +-
> >  lib/utils.c      |  1 +
> >  misc/ss.c        |  3 ++-
> >  tc/tc_class.c    |  2 +-
> >  tc/tc_filter.c   |  2 +-
> >  tc/tc_monitor.c  |  1 +
> >  tc/tc_qdisc.c    |  2 +-
> >  19 files changed, 44 insertions(+), 31 deletions(-)
> >   
> 
> no longer applies after merging main to next.
> 

I put the other version in iproute2 since wanted to test and it
did not depend on any later kernel changes