Re: [PATCH net-next v3 1/4] net: stmmac: Switch to zero-copy in non-XDP RX path

[Furong Xu <0x1207@gmail.com>]

On Thu, 23 Jan 2025 22:48:42 +0100, Andrew Lunn <andrew@lunn.ch> wrote:

> > Just to clarify, the patch that you had us try was not intended as an actual
> > fix, correct? It was only for diagnostic purposes, i.e. to see if there is
> > some kind of cache coherence issue, which seems to be the case?  So perhaps
> > the only fix needed is to add dma-coherent to our device tree?  
> 
> That sounds quite error prone. How many other DT blobs are missing the
> property? If the memory should be coherent, i would expect the driver
> to allocate coherent memory. Or the driver needs to handle
> non-coherent memory and add the necessary flush/invalidates etc.

stmmac driver does the necessary cache flush/invalidates to maintain cache lines
explicitly.

See dma_sync_single_for_cpu():
https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/tree/include/linux/dma-mapping.h#n297

dma_dev_need_sync() is supposed to return false for Tegra234, since the ethernet
controller on Tegra234 is dma coherent by SoC design as Brad said their
downstream device tree has dma-coherent turned on by default, and after add
dma-coherent to mainline ethernet node, stmmac driver works fine.
But dma-coherent property is missing in mainline Tegra234 ethernet device tree
node, dma_dev_need_sync() returns true and this is not the expected behavior.

The dma-coherent property in device tree node is SoC specific, so only the
vendors know if their stmmac ethernet controller is dma coherent and
whether their device tree are missing the critical dma-coherent property.