Re: [dpdk-dev] [PATCH v5 1/4] examples/multi_process/client_server_mp: check port validity

[Stephen Hemminger <stephen@networkplumber.org>]

On Wed, 7 Aug 2019 07:02:02 +0000
Matan Azrad <matan@mellanox.com> wrote:

> From: Stephen Hemminger
> > On Wed, 7 Aug 2019 05:38:42 +0000
> > Matan Azrad <matan@mellanox.com> wrote:
> >   
> > > From: Stephen Hemminger  
> > > > Sent: Wednesday, August 7, 2019 2:09 AM
> > > > To: Matan Azrad <matan@mellanox.com>
> > > > Cc: dev@dpdk.org; Stephen Hemminger <sthemmin@microsoft.com>
> > > > Subject: Re: [dpdk-dev] [PATCH v5 1/4]
> > > > examples/multi_process/client_server_mp: check port validity
> > > >
> > > > On Tue, 6 Aug 2019 20:03:22 +0000
> > > > Matan Azrad <matan@mellanox.com> wrote:
> > > >  
> > > > > >
> > > > > > The DPDK has lots of hard coded assumptions of all ports fitting in 64  
> > bits.  
> > > > > > Examples include testpmd/parameters.c etc.  
> > > > >
> > > > > Yes, I understand, but the user should know not to change the
> > > > > default value of RTE_MAX_ETHPORTS, at least it should be  
> > documented.  
> > > > >  
> > > > > > The original concept of a small set of assigned values for
> > > > > > portid is not going to scale. It really should have been more
> > > > > > like ifindex; something that is not used by common API's much
> > > > > > larger range; and  
> > > > assigned purely sequentially.  
> > > > > >
> > > > > > The API's should all be using names, but the DPDK port naming is
> > > > > > also a mess...  
> > > > >
> > > > > Port ID is OK, user can run port info, then to find the wanted
> > > > > port ID and  
> > > > configure it by port id list\bitmap.  
> > > > >  
> > > >
> > > >
> > > > The examples are toy programs. If user changes RTE_MAX_ETHPORTS it
> > > > will break lots of other places. Why put more checks in the
> > > > examples. Sorry, it really would not help to pretend that fixing the  
> > example is going to help this.  
> > >
> > >
> > > Agree that it is not needed to fix all the places now.
> > > It is better just to update the example documentation that  
> > RTE_MAX_ETHPORTS must not be changed when running this application.  
> > >
> > > I will ack your series(v7) , Consider to update the doc if you want to be  
> > completely perfect here.
> > 
> > Perhaps the right place to tell the users is somewhere in the documentation?
> > 
> > One place would be here:
> > 
> > diff --git a/doc/guides/faq/faq.rst b/doc/guides/faq/faq.rst index
> > f19c1389b6af..a847d9ceda22 100644
> > --- a/doc/guides/faq/faq.rst
> > +++ b/doc/guides/faq/faq.rst
> > @@ -195,3 +195,8 @@ Why can't my application receive packets on my
> > system with UEFI Secure Boot enab
> > 
> >  If UEFI secure boot is enabled, the Linux kernel may disallow the use of UIO
> > on the system.
> >  Therefore, devices for use by DPDK should be bound to the ``vfio-pci``
> > kernel module rather than ``igb_uio`` or ``uio_pci_generic``.
> > +
> > +What is the maximum number of ethernet devices?
> > +-----------------------------------------------
> > +
> > +The limit on the number of Ethernet devices is controlled by the
> > RTE_MAX_ETHPORTS configuration setting. Since many of the applications
> > use a 64 bit value for port mask; the current upper limit is 64 ports.
> >   
> I think there are systems with a lot of virtual ports which may use more than 64.
> 
> So update all the docs when the mask is defined, would be option too.

It would be good if (ie someone should do it but I don't have time);
to have a new type "port_set"  which is a variable length bit mask

It could be backwards compatible with existing usage.
Something like existing cpuset command format.

       Examples of the Mask Format:

           00000001                        # just bit 0 set
           40000000,00000000,00000000      # just bit 94 set
           00000001,00000000,00000000      # just bit 64 set
           000000ff,00000000               # bits 32-39 set
           00000000,000e3862               # 1,5,6,11-13,17-19 set