On 11 October 2018 at 14:28, Ilias Apalodimas
<ilias.apalodi...@linaro.org> wrote:
> commit 63ae7949e94a ("net: socionext: Use descriptor info instead of MMIO
> reads on Rx")
> removed constant mmio reads from the driver and started using a descriptor
> field to check if packet should be processed.
> This lead the napi rx handler being constantly called while no packets
> needed processing and ksoftirq getting 100% cpu usage. Issue one mmio read
> to clear the irq correcty after processing packets
>
> Signed-off-by: Ilias Apalodimas <ilias.apalodi...@linaro.org>
> Reported-by: Ard Biesheuvel <ard.biesheu...@linaro.org>
Tested-by: Ard Biesheuvel <ard.biesheu...@linaro.org>
Acked-by: Ard Biesheuvel <ard.biesheu...@linaro.org>
> ---
> drivers/net/ethernet/socionext/netsec.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/socionext/netsec.c
> b/drivers/net/ethernet/socionext/netsec.c
> index 7aa5ebb..4289ccb 100644
> --- a/drivers/net/ethernet/socionext/netsec.c
> +++ b/drivers/net/ethernet/socionext/netsec.c
> @@ -735,8 +735,11 @@ static int netsec_process_rx(struct netsec_priv *priv,
> int budget)
> u16 idx = dring->tail;
> struct netsec_de *de = dring->vaddr + (DESC_SZ * idx);
>
> - if (de->attr & (1U << NETSEC_RX_PKT_OWN_FIELD))
> + if (de->attr & (1U << NETSEC_RX_PKT_OWN_FIELD)) {
> + /* reading the register clears the irq */
> + netsec_read(priv, NETSEC_REG_NRM_RX_PKTCNT);
> break;
> + }
>
> /* This barrier is needed to keep us from reading
> * any other fields out of the netsec_de until we have
> --
> 2.7.4
>
