Re: [PATCH net-next] net: stmmac: qcom-ethqos: Add sysfs nodes for qcom ethqos

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Dec 06, 2023 at 05:17:25PM +0530, Sneh Shah wrote:
> 
> 
> On 12/5/2023 8:38 PM, Andrew Lunn wrote:
> > On Mon, Dec 04, 2023 at 02:18:54PM +0530, Sneh Shah wrote:
> >> Add sysfs nodes to conifigure routing of specific vlan id to GVM queue.
> >> GVM queue is not exposed to PVM stmmac, so TC ops can't configure routing.
> > 
> > Adding files in /sysfs has ~0 chance of being accepted.
> > 
> > As requested, please explain what all these different hardware blocks
> > are, and what you are trying to achieve. We can then recommend the
> > correct interface.
> > 
> >     Andrew
> > 
> > ---
> > pw-bot: cr
> 

> We have multiVM Architecture here. PVM will have stmmac running with
> 4 Rx Tx queues. stmmac in PVM is responsible to configure whole
> ethernet HW MAC/DMA/MTL ( including clocks, regulators and other
> core bsp elements).

Please remember that stmmac is mostly used in embedded systems. People
used to embedded systems generally don't know virtual machine
terminology. So please spell out what PBM, GVM, etc mean.

> In GVM we have thin Ethernet driver, which is responsible to
> configure and manage only 1 Rx/TX queue, i.e 5th Rx/Tx ethernet
> queue. GVM can't access any other resisters apart from this 5th
> queue specific MTL and DMA registers.
 
> We need to route vlan traffic of a specific Priority to GVM Queue
> (Ethernet queue 5) via programming a MAC register. The MAC register
> is not accessible in GVM and has to be programmed from PVM. stmmac
> already has TC OPS to program this routing via vlan
> priority. However, as PVM has only 4 queues enabled, TC tool will
> not take 5th queue as input. Hence, these nodes were added to
> conifure the MAC register to route specific vlan packets to 5th
> queue in GVM.
 
> Note: The queues mentioned above are HW MTL Queues and DMA
> Channels. The routing can be done in the HW itself based on vlan pcp
> before the packets reach to driver.

Is the normal way you would do this is like this:

tc qdisc add dev eth1 parent root handle 100 \
mqprio num_tc 4 \
map 0 1 2 3 0 0 0 0 0 0 0 0 0 0 0 0 \
queues 1@0 1@1 1@2 1@3 \
hw 1

But you are saying that you cannot extend this to 5 queues?

    Andrew




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [Linux for Sparc]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux