On 17.04.20 18:56, Philippe Schenker wrote:
On Fri, 2020-04-17 at 18:35 +0200, Stefano Babic wrote:
Hi Philippe,
On 11/03/20 11:52, Philippe Schenker wrote:
This patch enables the possibility to set FEC_ENET_ENABLE_TXC_DELAY or FEC_ENET_ENABLE_RXC_DELAY so one can via a define enable the RXC or TXC delay in the MAC.
Signed-off-by: Philippe Schenker philippe.schenker@toradex.com Reviewed-by: Oleksandr Suvorov oleksandr.suvorov@toradex.com
Changes in v3:
- Add Olek's reviewed-by tag
Changes in v2: None
drivers/net/fec_mxc.c | 10 ++++++++++ drivers/net/fec_mxc.h | 2 ++ 2 files changed, 12 insertions(+)
diff --git a/drivers/net/fec_mxc.c b/drivers/net/fec_mxc.c index bc5b63d7881..345d37be4e8 100644 --- a/drivers/net/fec_mxc.c +++ b/drivers/net/fec_mxc.c @@ -503,6 +503,16 @@ static int fec_open(struct eth_device *edev) writel(readl(&fec->eth->ecntrl) | FEC_ECNTRL_ETHER_EN, &fec->eth->ecntrl);
+#ifdef FEC_ENET_ENABLE_TXC_DELAY
- writel(readl(&fec->eth->ecntrl) | FEC_ECNTRL_TXC_DLY,
&fec->eth->ecntrl);+#endif
+#ifdef FEC_ENET_ENABLE_RXC_DELAY
- writel(readl(&fec->eth->ecntrl) | FEC_ECNTRL_RXC_DLY,
&fec->eth->ecntrl);+#endif
#if defined(CONFIG_MX25) || defined(CONFIG_MX53) || defined(CONFIG_MX6SL) udelay(100);
diff --git a/drivers/net/fec_mxc.h b/drivers/net/fec_mxc.h index 159aec89679..3c8fdda2638 100644 --- a/drivers/net/fec_mxc.h +++ b/drivers/net/fec_mxc.h @@ -188,6 +188,8 @@ struct ethernet_regs { #define FEC_ECNTRL_ETHER_EN 0x00000002 /* enable the FEC */ #define FEC_ECNTRL_SPEED 0x00000020 #define FEC_ECNTRL_DBSWAP 0x00000100 +#define FEC_ECNTRL_TXC_DLY 0x00010000 /* TXC delayed */ +#define FEC_ECNTRL_RXC_DLY 0x00020000 /* RXC delayed */
#define FEC_X_WMRK_STRFWD 0x00000100
But these are not exactly the properties removed by verdin-imx8mm in DTS, that is for example rgmii_rxc_dly ? And then, should not be better to support it in DT instead of a CONFIG_ ?
Hi Stefano,
I'll add Igor in cc as I'm not that familiar with most recent development on u-boot. Yes you are right. Those tags (rgmii_rxc_dly) are exactly the same. However they came in from our downstream branch and were never meant for mainline u-boot. We dropped them in favor of the proper phy-mode 'rgmii-id' in linux. Those originally probably also originated from our downstream linux branch...
But U-Boot does not support phy-modes on KSZ PHYs so I felt this was a straight forward and easy solution
This is an easy and straighforward solution, sure. Strange enough, your solution in DT is exactly as Oliver's did here:
http://patchwork.ozlabs.org/project/uboot/patch/20200203135619.10180-1-olive...
Both of you tried to get in in DT.
for u-boot that we can enable as soon as we have our new PHY KSZ9131 on our board. At the moment there is KSZ9031 placed which does indeed not support this.
Anyway, until the property is not mainlined to Linux, I am fine to merge your series.
Regards, Stefano