[dpdk-dev] [PKTGEN] [PATCH 1/2] usage_pktgen.rst: multiple instances: clean up section intro

[dpdk-dev] [PKTGEN] [PATCH 1/2] usage_pktgen.rst: multiple instances: clean up section intro

[Matthew Hall]

Signed-off-by: Matthew Hall <mhall@mhcomputing.net>
---
 docs/source/usage_pktgen.rst | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/docs/source/usage_pktgen.rst b/docs/source/usage_pktgen.rst
index 20bd314..efe8aa4 100644
--- a/docs/source/usage_pktgen.rst
+++ b/docs/source/usage_pktgen.rst
@@ -103,15 +103,15 @@ Multiple Instances of Pktgen or other application
 =================================================
 
 One possible solution I use and if you have enough ports available to use.
-Lets say you need two ports for your application, but you have 4 ports in
-your system. I physically loop back the cables to have port 0 connect to
-port 2 and port 1 connected to port 3. Now I can give two ports to my
-application and two ports to Pktgen.
-
-Setup if pktgen and your application you have to startup each one a bit
-differently to make sure they share the resources like memory and the
-ports. I will use two Pktgen running on the same machine, which just means
-you have to setup your application as one of the applications.
+Let's say you need two ports for your application, but you have 4 ports in
+your system. I physically loop back the cables to have port 0 connect to port
+2 and port 1 connected to port 3. Now I can give two ports to my application
+and two ports to Pktgen.
+
+If you are running pktgen and your application together, you have to start up
+each one a bit differently to make sure they share the resources like memory
+and the ports. I will use two Pktgens running on the same machine, which just
+means you have imagine your application as one of the applications.
 
 In my machine I have 8 10G ports and 72 lcores between 2 sockets. Plus I
 have 1024 hugepages per socket for a total of 2048.
-- 
2.5.0

[dpdk-dev] [PKTGEN] [PATCH 2/2] usage_pktgen.rst: multiple instances: clarify EAL options needed

[Matthew Hall]

Signed-off-by: Matthew Hall <mhall@mhcomputing.net>
---
 docs/source/usage_pktgen.rst | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/docs/source/usage_pktgen.rst b/docs/source/usage_pktgen.rst
index efe8aa4..223d033 100644
--- a/docs/source/usage_pktgen.rst
+++ b/docs/source/usage_pktgen.rst
@@ -157,4 +157,19 @@ The -m option then assigns lcores to the ports.
 The information from above is taken from two new files pktgen-master.sh
 and pktgen-slave.sh, have a look at them and adjust as you need.
 
+The following DPDK / EAL options must be configured correctly as well:
+
+* ``-l lcore_id_list``: non-conflicting list of lcores for each app
+
+* ``--master-lcore lcore_id``: non-conflicting master lcore for each app
+
+* ``-m hugepage_mb / --socket-mem hugepage_mb_list``: non-conflicting amount
+of hugepage memory for each app, or for each app on each CPU socket
+
+* ``--no-shconf``: prevents DPDK from claiming a lockfile that breaks
+concurrent use of multiple apps
+
+* ``--file-prefix``: assigns a unique name to the hugepage mmap() files for
+each app
+
 Pktgen can also be configured using the :ref:`commands`.
-- 
2.5.0

Re: [dpdk-dev] [PKTGEN] [PATCH 2/2] usage_pktgen.rst: multiple instances: clarify EAL options needed

[Wiles, Keith]

On 1/20/16, 11:50 PM, "dev on behalf of Matthew Hall" <dev-bounces@dpdk.org on behalf of mhall@mhcomputing.net> wrote:

>Signed-off-by: Matthew Hall <mhall@mhcomputing.net>
>---
> docs/source/usage_pktgen.rst | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
>diff --git a/docs/source/usage_pktgen.rst b/docs/source/usage_pktgen.rst
>index efe8aa4..223d033 100644
>--- a/docs/source/usage_pktgen.rst
>+++ b/docs/source/usage_pktgen.rst
>@@ -157,4 +157,19 @@ The -m option then assigns lcores to the ports.
> The information from above is taken from two new files pktgen-master.sh
> and pktgen-slave.sh, have a look at them and adjust as you need.
> 
>+The following DPDK / EAL options must be configured correctly as well:
>+
>+* ``-l lcore_id_list``: non-conflicting list of lcores for each app

If you are going to add -l option then you need to add the -c option too.
>+
>+* ``--master-lcore lcore_id``: non-conflicting master lcore for each app

This option is not required for pktgen or DPDK, which I think we can leave out.
>+
>+* ``-m hugepage_mb / --socket-mem hugepage_mb_list``: non-conflicting amount
>+of hugepage memory for each app, or for each app on each CPU socket
>+
>+* ``--no-shconf``: prevents DPDK from claiming a lockfile that breaks
>+concurrent use of multiple apps

This one is not required, correct? As long as you set the primary processes file-prefix to different prefixes for each one.
>+
>+* ``--file-prefix``: assigns a unique name to the hugepage mmap() files for
>+each app
>+
> Pktgen can also be configured using the :ref:`commands`.
>-- 
>2.5.0
>
>


Regards,
Keith

Re: [dpdk-dev] [PKTGEN] [PATCH 1/2] usage_pktgen.rst: multiple instances: clean up section intro

[Wiles, Keith]

On 1/20/16, 11:50 PM, "dev on behalf of Matthew Hall" <dev-bounces@dpdk.org on behalf of mhall@mhcomputing.net> wrote:

>Signed-off-by: Matthew Hall <mhall@mhcomputing.net>
>---
> docs/source/usage_pktgen.rst | 18 +++++++++---------
> 1 file changed, 9 insertions(+), 9 deletions(-)
>
>diff --git a/docs/source/usage_pktgen.rst b/docs/source/usage_pktgen.rst
>index 20bd314..efe8aa4 100644
>--- a/docs/source/usage_pktgen.rst
>+++ b/docs/source/usage_pktgen.rst
>@@ -103,15 +103,15 @@ Multiple Instances of Pktgen or other application
> =================================================
> 
> One possible solution I use and if you have enough ports available to use.
>-Lets say you need two ports for your application, but you have 4 ports in
>-your system. I physically loop back the cables to have port 0 connect to
>-port 2 and port 1 connected to port 3. Now I can give two ports to my
>-application and two ports to Pktgen.
>-
>-Setup if pktgen and your application you have to startup each one a bit
>-differently to make sure they share the resources like memory and the
>-ports. I will use two Pktgen running on the same machine, which just means
>-you have to setup your application as one of the applications.
>+Let's say you need two ports for your application, but you have 4 ports in
>+your system. I physically loop back the cables to have port 0 connect to port
>+2 and port 1 connected to port 3. Now I can give two ports to my application
>+and two ports to Pktgen.

It appears (if I compared the text correctly) the above only move a few trailing words to the next line, why?
>+
>+If you are running pktgen and your application together, you have to start up
>+each one a bit differently to make sure they share the resources like memory
>+and the ports. I will use two Pktgens running on the same machine, which just
>+means you have imagine your application as one of the applications.

Maybe this is clearer.

If you are running pktgen and your application together, you have to start up
each one a bit differently to make sure they share the resources like memory,
huge page files and ports. I will use two Pktgens running on the same machine,
which just means you have imagine your application as one of the Pktgen instances.


> 
> In my machine I have 8 10G ports and 72 lcores between 2 sockets. Plus I
> have 1024 hugepages per socket for a total of 2048.
>-- 
>2.5.0
>
>


Regards,
Keith

Re: [dpdk-dev] [PKTGEN] [PATCH 1/2] usage_pktgen.rst: multiple instances: clean up section intro

[Matthew Hall]

On Thu, Jan 21, 2016 at 03:46:21PM +0000, Wiles, Keith wrote:
> It appears (if I compared the text correctly) the above only move a few 
> trailing words to the next line, why?

I believe in trying to leave code / docs cleaner than I found them.

Most Markdown / ReStructured Text has a tradition of 78-character wide line 
length limits.

So just trying to make the world a better place. That's it.

> Maybe this is clearer.
> 
> If you are running pktgen and your application together, you have to start up
> each one a bit differently to make sure they share the resources like memory,
> huge page files and ports. I will use two Pktgens running on the same machine,
> which just means you have imagine your application as one of the Pktgen instances.

I'll make some adjustments when I get more time and try again.

In general I think it would be good if we had a bit more of a forest view of 
trying to make docs for all of the things pktgen docs are missing and not get 
too hung up on one tree or another.

I don't necessarily have a ton of time for editing these to the Nth degree, 
but I do want to make life better for the next confused people to avoid a 
storm of duplicate comments and issues.

Matthew.