RE: [patch net-next v2 0/3] net/sched: Improve getting objects by indexes

[Chris Mi]

> -Original Message-
> From: David Miller [mailto:da...@davemloft.net]
> Sent: Thursday, August 31, 2017 5:39 AM
> To: Chris Mi <chr...@mellanox.com>
> Cc: netdev@vger.kernel.org; j...@mojatatu.com;
> xiyou.wangc...@gmail.com; j...@resnulli.us; mawil...@microsoft.com
> Subject: Re: [patch net-next v2 0/3] net/sched: Improve getting objects by
> indexes
> 
> From: Chris Mi <chr...@mellanox.com>
> Date: Wed, 30 Aug 2017 02:31:56 -0400
> 
> > Using current TC code, it is very slow to insert a lot of rules.
> >
> > In order to improve the rules update rate in TC, we introduced the
> > following two changes:
> > 1) changed cls_flower to use IDR to manage the filters.
> > 2) changed all act_xxx modules to use IDR instead of
> >a small hash table
> >
> > But IDR has a limitation that it uses int. TC handle uses u32.
> > To make sure there is no regression, we add several new IDR APIs to
> > support unsigned long.
> >
> > v2
> > ==
> >
> > Addressed Hannes's comment:
> > express idr_alloc in terms of idr_alloc_ext and most of the other
> > functions
> 
> Series applied, thanks.

Thank you, David,

-Chris

Re: [patch net-next v2 0/3] net/sched: Improve getting objects by indexes

[David Miller]

From: Chris Mi 
Date: Wed, 30 Aug 2017 02:31:56 -0400

> Using current TC code, it is very slow to insert a lot of rules.
> 
> In order to improve the rules update rate in TC,
> we introduced the following two changes:
> 1) changed cls_flower to use IDR to manage the filters.
> 2) changed all act_xxx modules to use IDR instead of
>a small hash table
> 
> But IDR has a limitation that it uses int. TC handle uses u32.
> To make sure there is no regression, we add several new IDR APIs
> to support unsigned long.
> 
> v2
> ==
> 
> Addressed Hannes's comment:
> express idr_alloc in terms of idr_alloc_ext and most of the other functions

Series applied, thanks.

[patch net-next v2 0/3] net/sched: Improve getting objects by indexes

[Chris Mi]

Using current TC code, it is very slow to insert a lot of rules.

In order to improve the rules update rate in TC,
we introduced the following two changes:
1) changed cls_flower to use IDR to manage the filters.
2) changed all act_xxx modules to use IDR instead of
   a small hash table

But IDR has a limitation that it uses int. TC handle uses u32.
To make sure there is no regression, we add several new IDR APIs
to support unsigned long.

v2
==

Addressed Hannes's comment:
express idr_alloc in terms of idr_alloc_ext and most of the other functions

Chris Mi (3):
  idr: Add new APIs to support unsigned long
  net/sched: Change cls_flower to use IDR
  net/sched: Change act_api and act_xxx modules to use IDR

 include/linux/idr.h|  69 -
 include/linux/radix-tree.h |  21 +++-
 include/net/act_api.h  |  76 +-
 lib/idr.c  |  66 ++--
 lib/radix-tree.c   |   6 +-
 net/sched/act_api.c| 251 ++---
 net/sched/act_bpf.c|  17 ++-
 net/sched/act_connmark.c   |  16 ++-
 net/sched/act_csum.c   |  16 ++-
 net/sched/act_gact.c   |  16 ++-
 net/sched/act_ife.c|  20 ++--
 net/sched/act_ipt.c|  26 +++--
 net/sched/act_mirred.c |  19 ++--
 net/sched/act_nat.c|  16 ++-
 net/sched/act_pedit.c  |  18 ++--
 net/sched/act_police.c |  18 ++--
 net/sched/act_sample.c |  17 ++-
 net/sched/act_simple.c |  20 ++--
 net/sched/act_skbedit.c|  18 ++--
 net/sched/act_skbmod.c |  18 ++--
 net/sched/act_tunnel_key.c |  20 ++--
 net/sched/act_vlan.c   |  22 ++--
 net/sched/cls_flower.c |  55 +-
 23 files changed, 427 insertions(+), 414 deletions(-)

-- 
1.8.3.1