Re: [PATCH v2 net-next 1/2] net-sysfs: display two backlog queue len separately

[Jason Xing]

On Tue, Mar 14, 2023 at 10:59 PM Eric Dumazet <edumazet@xxxxxxxxxx> wrote:
>
> On Mon, Mar 13, 2023 at 8:06 PM Jason Xing <kerneljasonxing@xxxxxxxxx> wrote:
> >
> > From: Jason Xing <kernelxing@xxxxxxxxxxx>
> >
> > Sometimes we need to know which one of backlog queue can be exactly
> > long enough to cause some latency when debugging this part is needed.
> > Thus, we can then separate the display of both.
> >
> > Signed-off-by: Jason Xing <kernelxing@xxxxxxxxxxx>
> > ---
> > v2: keep the total len of backlog queues untouched as Eric said
> > Link: https://lore.kernel.org/lkml/20230311151756.83302-1-kerneljasonxing@xxxxxxxxx/
> > ---
> >  net/core/net-procfs.c | 20 ++++++++++++++++----
> >  1 file changed, 16 insertions(+), 4 deletions(-)
> >
> > diff --git a/net/core/net-procfs.c b/net/core/net-procfs.c
> > index 1ec23bf8b05c..2809b663e78d 100644
> > --- a/net/core/net-procfs.c
> > +++ b/net/core/net-procfs.c
> > @@ -115,10 +115,19 @@ static int dev_seq_show(struct seq_file *seq, void *v)
> >         return 0;
> >  }
> >
> > +static u32 softnet_input_pkt_queue_len(struct softnet_data *sd)
> > +{
> > +       return skb_queue_len_lockless(&sd->input_pkt_queue);
> > +}
> > +
> > +static u32 softnet_process_queue_len(struct softnet_data *sd)
> > +{
> > +       return skb_queue_len_lockless(&sd->process_queue);
> > +}
> > +
> >  static u32 softnet_backlog_len(struct softnet_data *sd)
> >  {
> > -       return skb_queue_len_lockless(&sd->input_pkt_queue) +
> > -              skb_queue_len_lockless(&sd->process_queue);
> > +       return softnet_input_pkt_queue_len(sd) + softnet_process_queue_len(sd);
>
[...]
> Reading these variables twice might lead to inconsistency that can
> easily be avoided.
>
> I would suggest you cache the values,
>
> u32 len1 = softnet_input_pkt_queue_len(sd);
> u32 len2 = softnet_process_queue_len(sd);

Agreed. Thank you, Eric. I should have realized that.

Also, the 2/2 patch which is all about the time_/budget_squeeze should
avoid such inconsistency, I think.

Jason
>
>
>
> >  }
> >
> >  static struct softnet_data *softnet_get_online(loff_t *pos)
> > @@ -169,12 +178,15 @@ static int softnet_seq_show(struct seq_file *seq, void *v)
> >          * mapping the data a specific CPU
> >          */
> >         seq_printf(seq,
> > -                  "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x\n",
> > +                  "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x "
> > +                  "%08x %08x\n",
> >                    sd->processed, sd->dropped, sd->time_squeeze, 0,
> >                    0, 0, 0, 0, /* was fastroute */
> >                    0,   /* was cpu_collision */
> >                    sd->received_rps, flow_limit_count,
> > -                  softnet_backlog_len(sd), (int)seq->index);
> > +                  softnet_backlog_len(sd),     /* keep it untouched */
>                     len1 + len2.
>
> > +                  (int)seq->index,
> > +                  softnet_input_pkt_queue_len(sd), softnet_process_queue_len(sd));
>                len1,  len2);
>
> >         return 0;
> >  }
> >
> > --
> > 2.37.3
> >