On Wed, Oct 23, 2024 at 03:08:52AM +0200, Andrew Lunn wrote: > On Wed, Oct 23, 2024 at 08:41:36AM +0800, Inochi Amaoto wrote: > > On Tue, Oct 22, 2024 at 03:51:08PM +0200, Andrew Lunn wrote: > > > On Tue, Oct 22, 2024 at 06:21:49PM +0800, Inochi Amaoto wrote: > > > > On Mon, Oct 21, 2024 at 03:27:18PM +0200, Andrew Lunn wrote: > > > > > > It is related to the RGMII delay. On sg2044, when the phy > > > > > > sets rx-delay, the interal mac is not set the same delay, > > > > > > so this is needed to be set. > > > > > > > > > > This is the wrong way to do it. Please look at how phy-mode should be > > > > > used, the four different "rgmii" values. Nearly everybody gets this > > > > > wrong, so there are plenty of emails from me in the netdev list about > > > > > how it should be done. > > > > > > > > > > > > > The phy-mode is alreay set to the "rgmii-id" and a rx delay is already > > > > set (a default tx delay is set by the phy driver). In the scenario > > > > the extra bit is used to fix 2ns difference between the sampling clock > > > > and data. It is more like an extra setting and the kernel can not handle > > > > it by only setting the phy-mode. > > > > > > This sounds wrong. > > > > > > So in DT you have rgmii-id? You say the PHY is doing TX delay. So you > > > pass PHY_INTERFACE_MODE_RGMII_TXID to the PHY? It is not clear from > > > this patch, i don't see any code mentioning > > > PHY_INTERFACE_MODE_RGMII_TXID. Could you point me at that code. > > > > > > Andrew > > > > The phy on the board I have is YT8531, The config code is here: > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/phy/motorcomm.c#n868 > > This PHY should be able to do rgmii-id, so there is no need for the > MAC to add delays. We encourage that setup in linux, so all RGMII > MAC/PHY pairs are the same, the PHY add the delays. > Yes, this is what I have done at the beginning. At first I only set up the phy setting and not set the config in the syscon. But I got a weird thing: the phy lookback test is timeout. Although the datasheet told it just adds a internal delay for the phy, I suspect sophgo does something more to set this delay. Regards, Inochi
