RE: [EXT] Re: [net-next] enetc: add support Credit Based Shaper(CBS) for hardware offload

[Po Liu]

Hi Jakub,

Thanks!


Br,
Po Liu

> -----Original Message-----
> From: Jakub Kicinski <jakub.kicinski@xxxxxxxxxxxxx>
> Sent: 2019年11月25日 7:35
> To: Po Liu <po.liu@xxxxxxx>
> Cc: davem@xxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> netdev@xxxxxxxxxxxxxxx; vinicius.gomes@xxxxxxxxx; Claudiu Manoil
> <claudiu.manoil@xxxxxxx>; Vladimir Oltean <vladimir.oltean@xxxxxxx>;
> Alexandru Marginean <alexandru.marginean@xxxxxxx>; Xiaoliang Yang
> <xiaoliang.yang_1@xxxxxxx>; Roy Zang <roy.zang@xxxxxxx>; Mingkai Hu
> <mingkai.hu@xxxxxxx>; Jerry Huang <jerry.huang@xxxxxxx>; Leo Li
> <leoyang.li@xxxxxxx>
> Subject: [EXT] Re: [net-next] enetc: add support Credit Based Shaper(CBS) for
> hardware offload
> 
> Caution: EXT Email
> 
> On Sat, 23 Nov 2019 19:02:09 -0800, Jakub Kicinski wrote:
> > On Fri, 22 Nov 2019 07:17:18 +0000, Po Liu wrote:
> > > +   if (tc == prio_top) {
> > > +           max_interference_size = port_frame_max_size * 8;
> > > +   } else {
> > > +           u32 m0, ma, r0, ra;
> > > +
> > > +           m0 = port_frame_max_size * 8;
> > > +           ma = enetc_port_rd(&si->hw, ENETC_PTCMSDUR(prio_top)) * 8;
> > > +           ra = enetc_get_cbs_bw(&si->hw, prio_top) *
> > > +                   port_transmit_rate * 10000ULL;
> > > +           r0 = port_transmit_rate * 1000000ULL;
> > > +           max_interference_size = m0 + ma + (u64)ra * m0 / (r0 - ra);
> > > +   }
> > > +
> > > +   /* hiCredit bits calculate by:
> > > +    *
> > > +    * maxSizedFrame * (idleSlope/portTxRate)
> > > +    */
> > > +   hi_credit_bit = max_interference_size * bw / 100;
> > > +
> > > +   /* hiCredit bits to hiCredit register need to calculated as:
> > > +    *
> > > +    * (enetClockFrequency / portTransmitRate) * 100
> > > +    */
> > > +   hi_credit_reg = (ENETC_CLK * 100ULL) * hi_credit_bit
> > > +                   / (port_transmit_rate * 1000000ULL);
> >
> > Hi! The patch looks good to me, but I'm concerned about those 64bit
> > divisions here. Don't these need to be div_u64() & co.? Otherwise we
> > may see one of the:
> >
> > ERROR: "__udivdi3" [drivers/net/ethernet/freescale/enetc/fsl-enetc.ko]
> undefined!
> >
> > messages from the build bot..
> >
> > I could be wrong, I haven't actually tested..
> 
> Yup:
> 
> drivers/net/ethernet/freescale/enetc/enetc_qos.o: In function
> `enetc_setup_tc_cbs':
> enetc_qos.c:(.text+0x5b4): undefined reference to `__udivdi3'
> enetc_qos.c:(.text+0x608): undefined reference to `__udivdi3'
> /home/jkicinski/devel/linux/Makefile:1077: recipe for target 'vmlinux' failed
> make[1]: *** [vmlinux] Error 1
> make[1]: Leaving directory '/home/jkicinski/devel/linux/build_tmp2'
> Makefile:179: recipe for target 'sub-make' failed
> make: *** [sub-make] Error 2
> 
> Please fix and repost.

Will update to div_u64().