Re: [dpdk-dev] [RFC] app/testpmd: support multi-process

[Wisam Monther <wisamm@nvidia.com>]

> -----Original Message-----
> From: oulijun <oulijun@huawei.com>
> Sent: Tuesday, January 12, 2021 4:13 PM
> To: Wisam Monther <wisamm@nvidia.com>; ferruh.yigit@intel.com;
> wenzhuo.lu@intel.com; beilei.xing@intel.com;
> bernard.iremonger@intel.com
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [RFC] app/testpmd: support multi-process
> 
> 
> 
> 在 2021/1/10 20:32, Wisam Monther 写道:
> > Hi,
> >
> >> -----Original Message-----
> >> From: dev <dev-bounces@dpdk.org> On Behalf Of Lijun Ou
> >> Sent: Friday, January 8, 2021 11:46 AM
> >> To: ferruh.yigit@intel.com; wenzhuo.lu@intel.com;
> >> beilei.xing@intel.com; bernard.iremonger@intel.com
> >> Cc: dev@dpdk.org
> >> Subject: [dpdk-dev] [RFC] app/testpmd: support multi-process
> >>
> >> This patch adds multi-process support for testpmd.
> >> The test cmd example as follows:
> >> the primary cmd:
> >> ./testpmd -w xxx --file-prefix=xx -l 0-1 -n 2 -- -i\
> >> --rxq=16 --txq=16 --num-procs=2 --proc-id=0 the secondary cmd:
> >> ./testpmd -w xxx --file-prefix=xx -l 2-3 -n 2 -- -i\
> >> --rxq=16 --txq=16 --num-procs=2 --proc-id=1
> >>
> >> Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
> >> Signed-off-by: Lijun Ou <oulijun@huawei.com>
> >> ---
> >>   app/test-pmd/cmdline.c    |   6 ++-
> >>   app/test-pmd/config.c     |   9 +++-
> >>   app/test-pmd/parameters.c |   9 ++++
> >>   app/test-pmd/testpmd.c    | 133
> ++++++++++++++++++++++++++++++++--
> >> ------------
> >>   app/test-pmd/testpmd.h    |   7 +++
> >>   5 files changed, 121 insertions(+), 43 deletions(-)
> >>
> >
> > +1 for having this support for testpmd.
> >
> > Some questions in my mind:
> > How are the queues distributing here? In example I see 16 defined, are
> they for one instance or for all?
> > Will all processes have same memory region? If installing one RTE_FLOW in
> one instance will be active for all?
> > Same question for detaching device in one instance, how it will reflect on
> others?
> > There is many other scenarios like this, how it will handle those?
> >
> Hi，Wisam Monther
>    Firstly， thank you for your questions. According to the current
> implementation scheme, all queues are evenly allocated to different
> processes based on proc_num and proc_id。
> The number of receiving queues, number of processes, and process ID are
> specified for the master and slave processes. After being created by the
> main process, the sending and receiving queues are evenly distributed to all
> processes. The following shows the calculation rule for the Testpmd to
> allocate queues to each process after the proc ID is specified.
> start(queue start id) = proc_id * nb_q / num_procs；
> end(queue end id) = start + nb_q / num_procs；
> 
> For example, if support 16 txq and rxq
> the 0 ~7 for primary process
> the 8 ~15 for secondary process
> 
> all process have the same memory region.

Ok great, Thanks that' make sense.
What about the impact of installing RTE_FLOW rule into one instance, does it apply on all? I'm talking about SW side, not HW side?
The impact of detaching device from secondary/primary only, how it's designed to handle such cases?

> 
> Thanks
> Lijun Ou
> 
> > BRs,
> > Wisam Jaddo
> > .
> >