Re: [PATCH] net: fec: limit register access on i.MX6UL

[Juergen Borleis]

Am Dienstag, dem 20.09.2022 um 14:46 +0200 schrieb Andrew Lunn:
> > +/* for i.MX6ul */
> > +static u32 fec_enet_register_offset_6ul[] = {
> > +       FEC_IEVENT, FEC_IMASK, FEC_R_DES_ACTIVE_0, FEC_X_DES_ACTIVE_0,
> > +       FEC_ECNTRL, FEC_MII_DATA, FEC_MII_SPEED, FEC_MIB_CTRLSTAT,
> > FEC_R_CNTRL,
> > +       FEC_X_CNTRL, FEC_ADDR_LOW, FEC_ADDR_HIGH, FEC_OPD, FEC_TXIC0,
> > FEC_RXIC0,
> > +       FEC_HASH_TABLE_HIGH, FEC_HASH_TABLE_LOW, FEC_GRP_HASH_TABLE_HIGH,
> > +       FEC_GRP_HASH_TABLE_LOW, FEC_X_WMRK, FEC_R_DES_START_0,
> > +       FEC_X_DES_START_0, FEC_R_BUFF_SIZE_0, FEC_R_FIFO_RSFL,
> > FEC_R_FIFO_RSEM,
> > +       FEC_R_FIFO_RAEM, FEC_R_FIFO_RAFL, FEC_RACC,
> > +       RMON_T_DROP, RMON_T_PACKETS, RMON_T_BC_PKT, RMON_T_MC_PKT,
> > +       RMON_T_CRC_ALIGN, RMON_T_UNDERSIZE, RMON_T_OVERSIZE, RMON_T_FRAG,
> > +       RMON_T_JAB, RMON_T_COL, RMON_T_P64, RMON_T_P65TO127,
> > RMON_T_P128TO255,
> > +       RMON_T_P256TO511, RMON_T_P512TO1023, RMON_T_P1024TO2047,
> > +       RMON_T_P_GTE2048, RMON_T_OCTETS,
> > +       IEEE_T_DROP, IEEE_T_FRAME_OK, IEEE_T_1COL, IEEE_T_MCOL, IEEE_T_DEF,
> > +       IEEE_T_LCOL, IEEE_T_EXCOL, IEEE_T_MACERR, IEEE_T_CSERR, IEEE_T_SQE,
> > +       IEEE_T_FDXFC, IEEE_T_OCTETS_OK,
> > +       RMON_R_PACKETS, RMON_R_BC_PKT, RMON_R_MC_PKT, RMON_R_CRC_ALIGN,
> > +       RMON_R_UNDERSIZE, RMON_R_OVERSIZE, RMON_R_FRAG, RMON_R_JAB,
> > +       RMON_R_RESVD_O, RMON_R_P64, RMON_R_P65TO127, RMON_R_P128TO255,
> > +       RMON_R_P256TO511, RMON_R_P512TO1023, RMON_R_P1024TO2047,
> > +       RMON_R_P_GTE2048, RMON_R_OCTETS,
> > +       IEEE_R_DROP, IEEE_R_FRAME_OK, IEEE_R_CRC, IEEE_R_ALIGN,
> > IEEE_R_MACERR,
> > +       IEEE_R_FDXFC, IEEE_R_OCTETS_OK
> > +};
> >  #else
> >  static __u32 fec_enet_register_version = 1;
> 
> Seeing this, i wonder if the i.MX6ul needs its own register version,
> so that ethtool(1) knows what registers are valid?

I don't think so. The register layout is the same in both SoCs, e.g. all
existing registers are at the same offsets on i.MX6 and i.MX6UL. And due to the
memset() call, the few missing registers on i.MX6UL are all reported as 0.

jb

-- 
Pengutronix e.K.                       | Juergen Borleis             |
Steuerwalder Str. 21                   | https://www.pengutronix.de/ |
31137 Hildesheim, Germany              | Phone: +49-5121-206917-128  |
Amtsgericht Hildesheim, HRA 2686       | Fax:   +49-5121-206917-9    |