[PATCHv2 net] Documentation: add description for net.core.gro_normal_batch

[PATCHv2 net] Documentation: add description for net.core.gro_normal_batch

[Xin Long]

Describe it in admin-guide/sysctl/net.rst like other Network core options.
Users need to know gro_normal_batch for performance tuning.

v1->v2:
  - Improved the description according to the suggestion from Edward and
    Jakub.

Fixes: 323ebb61e32b ("net: use listified RX for handling GRO_NORMAL skbs")
Reported-by: Prijesh Patel <prpatel@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
---
 Documentation/admin-guide/sysctl/net.rst | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/Documentation/admin-guide/sysctl/net.rst b/Documentation/admin-guide/sysctl/net.rst
index f86b5e1623c6..5cb99403bf03 100644
--- a/Documentation/admin-guide/sysctl/net.rst
+++ b/Documentation/admin-guide/sysctl/net.rst
@@ -374,6 +374,17 @@ option is set to SOCK_TXREHASH_DEFAULT (i. e. not overridden by setsockopt).
 If set to 1 (default), hash rethink is performed on listening socket.
 If set to 0, hash rethink is not performed.
 
+gro_normal_batch
+----------------
+
+Maximum number of the segments to batch up for GRO list-RX. When a packet exits
+GRO, either as a coalesced superframe or as an original packet which GRO has
+decided not to coalesce, it is placed on a per-NAPI list. This list is then
+passed to the stack when the segments in this list count towards the
+gro_normal_batch limit.
+
+Default : 8
+
 2. /proc/sys/net/unix - Parameters for Unix domain sockets
 ----------------------------------------------------------
 
-- 
2.31.1

Re: [PATCHv2 net] Documentation: add description for net.core.gro_normal_batch

[Jakub Kicinski]

On Fri, 13 May 2022 11:02:35 -0400 Xin Long wrote:
> Describe it in admin-guide/sysctl/net.rst like other Network core options.
> Users need to know gro_normal_batch for performance tuning.
> 
> v1->v2:
>   - Improved the description according to the suggestion from Edward and
>     Jakub.
> 
> Fixes: 323ebb61e32b ("net: use listified RX for handling GRO_NORMAL skbs")
> Reported-by: Prijesh Patel <prpatel@redhat.com>
> Signed-off-by: Xin Long <lucien.xin@gmail.com>
> ---
>  Documentation/admin-guide/sysctl/net.rst | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/Documentation/admin-guide/sysctl/net.rst b/Documentation/admin-guide/sysctl/net.rst
> index f86b5e1623c6..5cb99403bf03 100644
> --- a/Documentation/admin-guide/sysctl/net.rst
> +++ b/Documentation/admin-guide/sysctl/net.rst
> @@ -374,6 +374,17 @@ option is set to SOCK_TXREHASH_DEFAULT (i. e. not overridden by setsockopt).
>  If set to 1 (default), hash rethink is performed on listening socket.
>  If set to 0, hash rethink is not performed.
>  
> +gro_normal_batch
> +----------------
> +
> +Maximum number of the segments to batch up for GRO list-RX.

How about s/for GRO list-RX/on output of GRO/ ?

> When a packet exits
> +GRO, either as a coalesced superframe or as an original packet which GRO has
> +decided not to coalesce, it is placed on a per-NAPI list. This list is then
> +passed to the stack when the segments in this list count towards the
> +gro_normal_batch limit.

... when the number of segments reaches the gro_normal_batch limit.

> +
> +Default : 8

Also, should we drop the default? It's easy to grep for, chances are if
anyone updates the value they will forget to change the doc.

Sorry for the late review, I wasn't expecting v3 will be needed.

>  2. /proc/sys/net/unix - Parameters for Unix domain sockets
>  ----------------------------------------------------------
>  

Re: [PATCHv2 net] Documentation: add description for net.core.gro_normal_batch

[Xin Long]

On Tue, May 17, 2022 at 8:21 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Fri, 13 May 2022 11:02:35 -0400 Xin Long wrote:
> > Describe it in admin-guide/sysctl/net.rst like other Network core options.
> > Users need to know gro_normal_batch for performance tuning.
> >
> > v1->v2:
> >   - Improved the description according to the suggestion from Edward and
> >     Jakub.
> >
> > Fixes: 323ebb61e32b ("net: use listified RX for handling GRO_NORMAL skbs")
> > Reported-by: Prijesh Patel <prpatel@redhat.com>
> > Signed-off-by: Xin Long <lucien.xin@gmail.com>
> > ---
> >  Documentation/admin-guide/sysctl/net.rst | 11 +++++++++++
> >  1 file changed, 11 insertions(+)
> >
> > diff --git a/Documentation/admin-guide/sysctl/net.rst b/Documentation/admin-guide/sysctl/net.rst
> > index f86b5e1623c6..5cb99403bf03 100644
> > --- a/Documentation/admin-guide/sysctl/net.rst
> > +++ b/Documentation/admin-guide/sysctl/net.rst
> > @@ -374,6 +374,17 @@ option is set to SOCK_TXREHASH_DEFAULT (i. e. not overridden by setsockopt).
> >  If set to 1 (default), hash rethink is performed on listening socket.
> >  If set to 0, hash rethink is not performed.
> >
> > +gro_normal_batch
> > +----------------
> > +
> > +Maximum number of the segments to batch up for GRO list-RX.
>
> How about s/for GRO list-RX/on output of GRO/ ?
>
> > When a packet exits
> > +GRO, either as a coalesced superframe or as an original packet which GRO has
> > +decided not to coalesce, it is placed on a per-NAPI list. This list is then
> > +passed to the stack when the segments in this list count towards the
> > +gro_normal_batch limit.
>
> ... when the number of segments reaches the gro_normal_batch limit.
>
> > +
> > +Default : 8
>
> Also, should we drop the default? It's easy to grep for, chances are if
> anyone updates the value they will forget to change the doc.
>
> Sorry for the late review, I wasn't expecting v3 will be needed.
No worries, v3 has been posted, please check.

Thanks.

>
> >  2. /proc/sys/net/unix - Parameters for Unix domain sockets
> >  ----------------------------------------------------------
> >
>