[dpdk-dev] standardize device identification

Thomas Monjalon thomas at monjalon.net
Fri Jan 5 16:34:03 CET 2018


05/01/2018 15:14, Finn Christensen:
> From: Thomas Monjalon [mailto:thomas at monjalon.net]
> >05/01/2018 12:09, Finn Christensen:
> >> From: Thomas Monjalon
> >>     Which property can help to distinguish Napatech ports?
> >>     Can you use class=eth,dev_port=X ?
> >>     The dev_port property will use /sys/class/net/DEV/dev_port on Linux. Is it
> >>     OK for you?
> >>
> >> Actually, what we were thinking of was using the mac property in the
> >> class category to distinguish our ports.
> >> For instance:
> >> -w bus=pci,id=0000:01:00.0/class=eth,mac=00:11:22:33:44:55
> >> or simply:
> >> -w class=eth,mac=00:11:22:33:44:55
> >
> >The problem with the mac property is that it cannot be used for
> >white/blacklisting in DPDK because the MAC is not known before port
> >initialization.
> >
> 
> Sure, that makes sense. I just for a minute thought that we could use that
> mechanism to enable individual ports at startup also. We will continue to
> use proprietary devargs passed by whiterlist to the PMD probe function.
> What we needed was a way to select the individual ports, by using
> rte_eth_dev_get_port_by_name().

The whitelist will be replaced by this new syntax.
And yes, you can have your own driver-specific property with this syntax.

> >> We will not be able to support the dev_port property, that will not work for
> >us.
> >> At least not for now.

It leads to a totally different question:
Can we discuss again how to integrate your driver in DPDK upstream?
Please explain again your situation in a new thread.


More information about the dev mailing list