From: Eric Dumazet
Date: Wed, 21 Feb 2018 06:43:03 -0800
> From: Eric Dumazet
>
> BBR uses tcp_tso_autosize() in an attempt to probe what would be the
> burst sizes and to adjust cwnd in bbr_target_cwnd() with following
> gold formula :
>
> /* Allow enough full-sized skbs in flight to utilize
On Wed, 2018-02-21 at 07:09 -0800, Eric Dumazet wrote:
> On Wed, Feb 21, 2018 at 7:01 AM, Paolo Abeni wrote:
> >
> > Very minor nit, why don't:
> >
> > return max_t(u32, bytes / mss_now, min_tso_segs);
> >
> > and drop the 'segs' local variable?
>
> Simply to ease backports.
>
> We ha
On Wed, Feb 21, 2018 at 10:14 AM Neal Cardwell wrote:
> On Wed, Feb 21, 2018 at 9:43 AM, Eric Dumazet
wrote:
> >
> > From: Eric Dumazet
> >
> > BBR uses tcp_tso_autosize() in an attempt to probe what would be the
> > burst sizes and to adjust cwnd in bbr_target_cwnd() with following
> > gold fo
On Wed, Feb 21, 2018 at 9:43 AM, Eric Dumazet wrote:
>
> From: Eric Dumazet
>
> BBR uses tcp_tso_autosize() in an attempt to probe what would be the
> burst sizes and to adjust cwnd in bbr_target_cwnd() with following
> gold formula :
>
> /* Allow enough full-sized skbs in flight to utilize end s
On Wed, Feb 21, 2018 at 7:01 AM, Paolo Abeni wrote:
>
> Very minor nit, why don't:
>
> return max_t(u32, bytes / mss_now, min_tso_segs);
>
> and drop the 'segs' local variable?
Simply to ease backports.
We had some constant changes in this function in the past.
Hi,
On Wed, 2018-02-21 at 06:43 -0800, Eric Dumazet wrote:
> From: Eric Dumazet
>
> BBR uses tcp_tso_autosize() in an attempt to probe what would be the
> burst sizes and to adjust cwnd in bbr_target_cwnd() with following
> gold formula :
>
> /* Allow enough full-sized skbs in flight to utilize
From: Eric Dumazet
BBR uses tcp_tso_autosize() in an attempt to probe what would be the
burst sizes and to adjust cwnd in bbr_target_cwnd() with following
gold formula :
/* Allow enough full-sized skbs in flight to utilize end systems. */
cwnd += 3 * bbr->tso_segs_goal;
But GSO can be lacking o
