Re: [PATCH net v2] net: stmmac: Clear DMA errors for multi-channel interrupts

[Magnus Leatti]

> Since RBU and RPS were enabled only for statistics, was masking them in
> multi-MSI mode considered as an alternative?

Hi Maxime,

Tested that route on dwmac4 (YT6801, 7.3.3) next to v2.

Stock 7.3.3 dies in 4s under a 4x 64B UDP flood (122k irq/s storm).
Both v2 and dropping RBU from the abnormal mask survive with the same
~3M RX frames within noise (3 runs each, alternating), but:

                 v2 (A)        no RBU irq (B)
  MAC irqs:      18k - 43k     0
  rx_buf_unav:   9k - 21k      0
  %irq CPU:      3.9 - 7.1     1.1 - 1.5
  RX frames:     2.5M - 3.0M   2.8M - 3.2M

Same on 8x UDP + TCP TX (30s): both stay up, A takes up to 33k MAC
irqs, B takes 0.

Caveats for B:
- Core is a 5.2, so only the 4.10 mask was exercised.
- RPS never fired, so this doesn't test the RPS storm fix from v2.
- rx_buf_unav_irq stays at 0 (pre-1b9707e6f1a9).

Want RBU dropped from the dwmac4 mask too? Can send a patch or you can
fold it into v3.

Magnus