[ovs-dev] [v4 02/12] dpif-netdev: Add auto validation function for miniflow extract

Stokes, Ian ian.stokes at intel.com
Tue Jun 29 16:46:48 UTC 2021


> > -----Original Message-----
> > From: Amber, Kumar <kumar.amber at intel.com>
> > Sent: Thursday, June 24, 2021 12:27 PM
> > To: Stokes, Ian <ian.stokes at intel.com>; dev at openvswitch.org; Van Haaren,
> Harry
> > <harry.van.haaren at intel.com>
> > Cc: i.maximets at ovn.org
> > Subject: RE: [ovs-dev] [v4 02/12] dpif-netdev: Add auto validation function for
> > miniflow extract
> 
> <snip lots of patch>
> 
> > > >  #endif /* DPIF_NETDEV_AVX512_EXTRACT */ diff --git
> > > > a/lib/dpif-netdev.c b/lib/dpif-netdev.c index 567ebd952..4f4ab2790
> > > > 100644
> > > > --- a/lib/dpif-netdev.c
> > > > +++ b/lib/dpif-netdev.c
> > > > @@ -1181,8 +1181,8 @@ dpif_miniflow_extract_impl_set(struct
> > > > unixctl_conn *conn, int argc,
> > > >      struct ds reply = DS_EMPTY_INITIALIZER;
> > > >      ds_put_format(&reply, "Miniflow implementation set to %s.\n",
> > > mfex_name);
> > > >      const char *reply_str = ds_cstr(&reply);
> > > > -    unixctl_command_reply(conn, reply_str);
> > > >      VLOG_INFO("%s", reply_str);
> > > > +    unixctl_command_reply(conn, reply_str);
> > >
> > > Is there a reason for swapping the order above?
> > >
> >
> > This looks more logical .
> 
> Hi All,
> 
> Actually yes there's a good reason, by logging internally in Vswitchd first,
> and then sending the reply to the user, the order of prints in the logs is
> easier to understand.
> 
> This is particularly true when e.g. MFEX enabling logs can come *after* the PMD
> log
> print of study having chosen a specific MFEX impl.
> 
> (pseudo) Example of bad logging behaviour:
> <ovs startup stuff>
> PMD: MFEX study chose profile "eth_ip_udp" (128/128 hits)
> DPIF: MFEX optimized functionality set to "study"
> 
> (pseudo) Example of good logging behaviour (with switched log/conn_reply):
> <ovs startup stuff>
> DPIF: MFEX optimized functionality set to "study"
> PMD: MFEX study chose profile "eth_ip_udp" (128/128 hits)
> 
> Hope the helps clarify the change! -Harry

Thanks Harry, that makes it clearer for sure. Is it worth detailing the change in behaviour here that it does not get reverted in future by accident?

Regards
Ian


More information about the dev mailing list