mbox series

[v3,0/1] gro: decrease size of CB

Message ID 20230601160924.GA9194@debian
Headers
Series gro: decrease size of CB |

Message

Richard Gobert June 1, 2023, 4:09 p.m. UTC 
  This patch frees up space in the GRO CB, which is currently at its maximum
size. This patch was submitted and reviewed previously in a patch series,
but is now reposted as a standalone patch, as suggested by Paolo.
(https://lore.kernel.org/netdev/889f2dc5e646992033e0d9b0951d5a42f1907e07.camel@redhat.com/)

Changelog:

v2 -> v3:
  * add comment

v1 -> v2:
  * remove inline keyword

Richard Gobert (1):
  gro: decrease size of CB

 include/net/gro.h | 26 ++++++++++++++++----------
 net/core/gro.c    | 19 ++++++++++++-------
 2 files changed, 28 insertions(+), 17 deletions(-)

Comments

Alexander Lobakin June 2, 2023, 2:22 p.m. UTC | #1 
From: Richard Gobert <richardbgobert@gmail.com>
Date: Thu, 1 Jun 2023 18:09:28 +0200

> This patch frees up space in the GRO CB, which is currently at its maximum
> size. This patch was submitted and reviewed previously in a patch series,
> but is now reposted as a standalone patch, as suggested by Paolo.
> (https://lore.kernel.org/netdev/889f2dc5e646992033e0d9b0951d5a42f1907e07.camel@redhat.com/)
> 
> Changelog:
> 
> v2 -> v3:
>   * add comment
> 
> v1 -> v2:
>   * remove inline keyword

I hope you've checked that there's no difference in object code with and
w/o `inline`? Sometimes the compilers do weird things and stop inlining
oneliners if they're used more than once. skb_gro_reset_offset() is
marked `inline` exactly due to that =\

> 
> Richard Gobert (1):
>   gro: decrease size of CB
> 
>  include/net/gro.h | 26 ++++++++++++++++----------
>  net/core/gro.c    | 19 ++++++++++++-------
>  2 files changed, 28 insertions(+), 17 deletions(-)
Thanks,
Olek
Richard Gobert June 5, 2023, 1:58 p.m. UTC | #2 
> I hope you've checked that there's no difference in object code with and
> w/o `inline`? Sometimes the compilers do weird things and stop inlining
> oneliners if they're used more than once. skb_gro_reset_offset() is
> marked `inline` exactly due to that =\

Checked on standard x86-64 and arm64 gcc compilers.
Would you check any other cases?
patchwork-bot+netdevbpf@kernel.org June 6, 2023, 9:20 a.m. UTC | #3 
Hello:

This patch was applied to netdev/net-next.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Thu, 1 Jun 2023 18:09:28 +0200 you wrote:
> This patch frees up space in the GRO CB, which is currently at its maximum
> size. This patch was submitted and reviewed previously in a patch series,
> but is now reposted as a standalone patch, as suggested by Paolo.
> (https://lore.kernel.org/netdev/889f2dc5e646992033e0d9b0951d5a42f1907e07.camel@redhat.com/)
> 
> Changelog:
> 
> [...]

Here is the summary with links:
  - [v3,1/1] gro: decrease size of CB
    https://git.kernel.org/netdev/net-next/c/7b355b76e2b3

You are awesome, thank you!
Alexander Lobakin June 6, 2023, 1:24 p.m. UTC | #4 
From: Richard Gobert <richardbgobert@gmail.com>
Date: Mon, 5 Jun 2023 15:58:23 +0200

>> I hope you've checked that there's no difference in object code with and
>> w/o `inline`? Sometimes the compilers do weird things and stop inlining
>> oneliners if they're used more than once. skb_gro_reset_offset() is
>> marked `inline` exactly due to that =\
> 
> Checked on standard x86-64 and arm64 gcc compilers.
> Would you check any other cases?
I'd say it's enough, so no problem. Sometimes some odd things happen,
but you can never predict all of them.
I can't give Reviewed-by now that it has been already applied :D Had a
long weekend and all that. But the change is really good.

Thanks,
Olek