Re: [PATCH net 2/3] net: stmmac: dwmac1000: don't discard frames with only a checksum error

[opensource]

> Can this mask let giant frames through when Rx checksum offload is
> disabled?

Yes, in threshold Rx mode. The dangerous case was a giant frame that
spans two buffers, because stmmac_rx() parsed its last descriptor as a
new frame. The same happened without this series for any frame larger
than the Rx buffer.

Commit 834ec9530832 ("net: stmmac: drop split frames on AF_XDP
zero-copy RX") has since fixed that in net, and v2 will be based on it.

>
> Should bit 7 also count as a discard reason unless IPC (hw->rx_csum) is
> enabled?

Bit 7 alone does not identify a giant frame, because on cores with the
advanced timestamp feature it reports a timestamp instead.
With the current code enh_desc_get_rx_status() does not know which core
it runs on, and it has no access to hw->rx_csum.

A giant frame that fits one buffer therefore reaches the stack with a
verified FCS in threshold Rx mode only.

Dropping it would need a new parameter that tells the parser whether
the core has the timestamp feature.