On Fri. 27 May 2022 at 00:51, Srinivas Neeli <sneeli@xxxxxxxxxx> wrote:
> Hi Vincent,
>
> > -----Original Message-----
> > From: Vincent Mailhol <vincent.mailhol@xxxxxxxxx>
> > Sent: Friday, May 13, 2022 6:54 AM
> > To: Srinivas Neeli <sneeli@xxxxxxxxxx>
> > Cc: wg@xxxxxxxxxxxxxx; mkl@xxxxxxxxxxxxxx; davem@xxxxxxxxxxxxx;
> > edumazet@xxxxxxxxxx; Appana Durga Kedareswara Rao
> > <appanad@xxxxxxxxxx>; Srinivas Goud <sgoud@xxxxxxxxxx>; Michal Simek
> > <michals@xxxxxxxxxx>; kuba@xxxxxxxxxx; pabeni@xxxxxxxxxx; linux-
> > can@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; linux-arm-
> > kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; git
> > <git@xxxxxxxxxx>
> > Subject: Re: [PATCH] can: xilinx_can: Add Transmitter delay compensation
> > (TDC) feature support
> >
> > On Fri. 13 May 2022 at 07:30, Srinivas Neeli <srinivas.neeli@xxxxxxxxxx>
> > wrote:
> > > Added Transmitter delay compensation (TDC) feature support.
> > > In the case of higher measured loop delay with higher baud rates,
> > > observed bit stuff errors.
> > > By enabling the TDC feature in a controller, will compensate for the
> > > measure loop delay in the receive path.
> > > TDC feature requires BRP values can be 1 or 2.
> > > The current CAN framework does not limit the brp so while using TDC,
> > > have to restrict BRP values.
> > > Ex:
> > > ip link set can0 type can tq 12 prop-seg 39 phase-seg1 20 phase-seg2
> > > 20 sjw 20 dtq 12 dprop-seg 5 dphase-seg1 6 dphase-seg2 4 dsjw 4 fd on
> > > loopback on tdco 12 tdc-mode auto
> >
> > Did you experience some cases in which you had BRP > 2 and saw
> > transmission errors due to the absence of delay compensation? Could you
> > show the calculated values?
> > Usually, you start to observe but stuff error at high bitrates (e.g.
> > ~5MBPS), and for such bitrates, time quanta has to be small which then
> > results in a small BRP.
>
> yes, we observed errors with higher baud rates(4 and 5 MBPS).
> Observation:
> BRPA 1Mbps Sampling 75%
> BRPD 5MBPS Sampling 75%
> On NXP PHY observed a delay of 160 ns. so observing the failure.
> After enabling the TDC feature to work fine.
Can you also share the results of:
| ip --details link show can0
for both the automatic calculation by the CAN framework and for your
hand calculated values?
Thank you!
> > > Signed-off-by: Srinivas Neeli <srinivas.neeli@xxxxxxxxxx>
> > > ---
> > > drivers/net/can/xilinx_can.c | 30 +++++++++++++++++++++++++-----
> > > 1 file changed, 25 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/drivers/net/can/xilinx_can.c
> > > b/drivers/net/can/xilinx_can.c index e2b15d29d15e..7af518fbed02 100644
> > > --- a/drivers/net/can/xilinx_can.c
> > > +++ b/drivers/net/can/xilinx_can.c
> > > @@ -1,7 +1,7 @@
> > > // SPDX-License-Identifier: GPL-2.0-or-later
> > > /* Xilinx CAN device driver
> > > *
> > > - * Copyright (C) 2012 - 2014 Xilinx, Inc.
> > > + * Copyright (C) 2012 - 2022 Xilinx, Inc.
> > > * Copyright (C) 2009 PetaLogix. All rights reserved.
> > > * Copyright (C) 2017 - 2018 Sandvik Mining and Construction Oy
> > > *
> > > @@ -133,6 +133,8 @@ enum xcan_reg {
> > > #define XCAN_DLCR_BRS_MASK 0x04000000 /* BRS Mask in DLC */
> > >
> > > /* CAN register bit shift - XCAN_<REG>_<BIT>_SHIFT */
> > > +#define XCAN_BRPR_TDCO_SHIFT_CANFD 8 /* Transmitter Delay
> > Compensation Offset */
> >
> > Having CANFD in the name is redundant (TDC implies CANFD).
> > #define XCAN_BRPR_TDCO_SHIFT 8
> update in V2.
>
> >
> > > +#define XCAN_BRPR_TDCE_SHIFT_CANFD 16 /* Transmitter Delay
> > Compensation (TDC) Enable */
> >
> > Why not:
> > #define XCAN_BRPR_TDC_ENABLE BIT(16)
> update in V2.
>
> >
> > > #define XCAN_BTR_SJW_SHIFT 7 /* Synchronous jump width */
> > > #define XCAN_BTR_TS2_SHIFT 4 /* Time segment 2 */
> > > #define XCAN_BTR_SJW_SHIFT_CANFD 16 /* Synchronous jump width
> > */
> > > @@ -259,7 +261,7 @@ static const struct can_bittiming_const
> > xcan_bittiming_const_canfd2 = {
> > > .tseg2_min = 1,
> > > .tseg2_max = 128,
> > > .sjw_max = 128,
> > > - .brp_min = 2,
> > > + .brp_min = 1,
> >
> > Was there any reason to have brp_min = 2 in the first place?
> > I think this change should be a different patch. If the brp_min = 2 is just a
> > typo, you might also want to backport it to stable branches.
>
> On early silicon engineering samples we observed bit shrinking issue when we use brp =1 , hence we updated brp_min =2.
> As in production silicon this issue is fixed we are planning to revert the patch.
Great!
> > > .brp_max = 256,
> > > .brp_inc = 1,
> > > };
> > > @@ -272,11 +274,21 @@ static struct can_bittiming_const
> > xcan_data_bittiming_const_canfd2 = {
> > > .tseg2_min = 1,
> > > .tseg2_max = 16,
> > > .sjw_max = 16,
> > > - .brp_min = 2,
> > > + .brp_min = 1,
> > > .brp_max = 256,
> > > .brp_inc = 1,
> > > };
> > >
> > > +/* Transmission Delay Compensation constants for CANFD2.0 and Versal
> > > +*/ static const struct can_tdc_const xcan_tdc_const = {
> > > + .tdcv_min = 0,
> > > + .tdcv_max = 0, /* Manual mode not supported. */
> >
> > Right, had a look at the datasheet and xilinx indeed does not support setting
> > TDCV.
> > However, xilinx still has a TDCV register to report the measured transmission
> > delay.
> >
> > Socket CAN's TDC framework provides can_priv::do_get_auto_tdcv() to
> > report the measured value through the netlink interface:
> > https://elixir.bootlin.com/linux/v5.17/source/include/linux/can/dev.h#L87
> >
> > Can you implement this call back function?
> Will implement in V2.
>
> >
> > > + .tdco_min = 0,
> > > + .tdco_max = 64,
> > > + .tdcf_min = 0, /* Filter window not supported */
> > > + .tdcf_max = 0,
> > > +};
> > > +
> > > /**
> > > * xcan_write_reg_le - Write a value to the device register little endian
> > > * @priv: Driver private data structure
> > > @@ -425,6 +437,11 @@ static int xcan_set_bittiming(struct net_device
> > *ndev)
> > > priv->devtype.cantype == XAXI_CANFD_2_0) {
> > > /* Setting Baud Rate prescalar value in F_BRPR Register */
> > > btr0 = dbt->brp - 1;
> > > + if (can_tdc_is_enabled(&priv->can)) {
> > > + btr0 = btr0 |
> > > + (priv->can.tdc.tdco) << XCAN_BRPR_TDCO_SHIFT_CANFD |
> > > + 1 << XCAN_BRPR_TDCE_SHIFT_CANFD;
> >
> > I don't think the parenthesis around (priv->can.tdc.tdco) are needed.
> Yes, will update.
> >
> > btr0 = btr0 |
> > priv->can.tdc.tdco << XCAN_BRPR_TDCO_SHIFT |
> > XCAN_BRPR_TDC_ENABLE
> >
> > (c.f. above for macro names)
> >
> > > + }
> > >
> > > /* Setting Time Segment 1 in BTR Register */
> > > btr1 = dbt->prop_seg + dbt->phase_seg1 - 1; @@
> > > -1747,13 +1764,16 @@ static int xcan_probe(struct platform_device *pdev)
> > > priv->can.data_bittiming_const =
> > > &xcan_data_bittiming_const_canfd;
> > >
> > > - if (devtype->cantype == XAXI_CANFD_2_0)
> > > + if (devtype->cantype == XAXI_CANFD_2_0) {
> > > priv->can.data_bittiming_const =
> > > &xcan_data_bittiming_const_canfd2;
> > > + priv->can.tdc_const = &xcan_tdc_const;
> > > + }
> > >
> > > if (devtype->cantype == XAXI_CANFD ||
> > > devtype->cantype == XAXI_CANFD_2_0)
> > > - priv->can.ctrlmode_supported |= CAN_CTRLMODE_FD;
> > > + priv->can.ctrlmode_supported |= CAN_CTRLMODE_FD |
> > > + CAN_CTRLMODE_TDC_AUTO;
> > >
> > > priv->reg_base = addr;
> > > priv->tx_max = tx_max;
