sun8i_dwmac_reset() asserts EMAC_BASIC_CTL1.SOFT_RST and polls for the hardware to clear it. That bit only clears once the MAC has a running receive clock, which on external-PHY boards is driven by the PHY. Bringing the interface down powers the PHY down: phy_detach() calls phy_suspend(), which for a PHY without wake-on-LAN ends in BMCR_PDOWN. Boards that wire no reset line to their PHY, and share its supply with other always-on consumers, have nothing that undoes that. On the Orange Pi Zero 3 the Motorcomm YT8531 reset is, in the words of the board's upstream author, "hardwired via a simple RC circuit, so there is no GPIO", and phy-supply points at a regulator-always-on rail shared with four GPIO banks and the SD card. So after a warm reboot the PHY comes back still powered down. Probe asserts SOFT_RST, no receive clock arrives, and the reset never completes. Read off an affected board at boot, before anything touched the PHY: BMCR 0x1800 (PDOWN set) and EMAC_BASIC_CTL1 0x08000001, still set after polling for ten seconds -- so raising the 100 ms timeout does not help. Probe fails with -ETIMEDOUT and the interface never appears. That the interface teardown is what does it can be shown directly. A reboot via sysrq-b, which skips both the ifdown and device_shutdown(), comes up with PDOWN clear and resets fine; taking the interface down first and then using sysrq-b -- so the driver's own shutdown path still never runs -- reproduces the failure. The driver already has the right place for the reset. sun8i_dwmac_dma_reset() is registered as stmmac_dma_ops->reset and is documented as "reset the EMAC", but only zeroes a few registers. stmmac calls it from stmmac_init_dma_engine(), under stmmac_hw_setup(), whose only two callers are __stmmac_open() and stmmac_resume() -- both of which run after stmmac_init_phy() has attached and resumed the PHY. Doing the soft reset there means the receive clock is running by construction, for every PHY, whether or not a PHY driver is bound, without the MAC driver reaching into phylib. sun8i is the only stmmac variant that soft-resets at probe rather than in the reset hook; this brings it into line with the others. sun8i_dwmac_reset() stays for the mdio-mux switch callback, which needs a reset after changing the syscon and cannot use the hook. Tested on an Orange Pi Zero 3 (H618, YT8531, rgmii-rxid). Five warm reboots: no EMAC reset timeout, interface up at 1Gbps each time. One cold boot with power physically cycled, to cover the path this moves for boards that never hit the bug: same result. 20000 and 5000 1472-byte frames respectively, no loss, every MAC error counter at zero. Fixes: 9f93ac8d4085 ("net-next: stmmac: Add dwmac-sun8i") Signed-off-by: Pedro Santos --- .../net/ethernet/stmicro/stmmac/dwmac-sun8i.c | 49 ++++++++++++------- 1 file changed, 30 insertions(+), 19 deletions(-) diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c index 48c52eb..748ebab 100644 --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c @@ -272,8 +272,34 @@ static const struct emac_variant emac_variant_h6 = { /* sun8i_dwmac_dma_reset() - reset the EMAC * Called from stmmac via stmmac_dma_ops->reset */ +static int sun8i_dwmac_soft_reset(void __iomem *ioaddr) +{ + u32 v; + + v = readl(ioaddr + EMAC_BASIC_CTL1); + writel(v | 0x01, ioaddr + EMAC_BASIC_CTL1); + + /* The timeout was previously set to 10ms, but some board (OrangePI0) + * need more if no cable plugged. 100ms seems OK + */ + return readl_poll_timeout(ioaddr + EMAC_BASIC_CTL1, v, + !(v & 0x01), 100, 100000); +} + static int sun8i_dwmac_dma_reset(void __iomem *ioaddr) { + int err; + + /* The MAC soft reset only completes once the PHY is driving the RX + * clock. Doing it here rather than at probe means phylib has already + * attached and resumed the PHY, so the clock is running by + * construction -- including after a warm reboot that left the PHY + * powered down. + */ + err = sun8i_dwmac_soft_reset(ioaddr); + if (err) + return err; + writel(0, ioaddr + EMAC_RX_CTL1); writel(0, ioaddr + EMAC_TX_CTL1); writel(0, ioaddr + EMAC_RX_FRM_FLT); @@ -740,23 +766,12 @@ static void sun8i_dwmac_flow_ctrl(struct mac_device_info *hw, static int sun8i_dwmac_reset(struct stmmac_priv *priv) { - u32 v; - int err; - - v = readl(priv->ioaddr + EMAC_BASIC_CTL1); - writel(v | 0x01, priv->ioaddr + EMAC_BASIC_CTL1); + int err = sun8i_dwmac_soft_reset(priv->ioaddr); - /* The timeout was previously set to 10ms, but some board (OrangePI0) - * need more if no cable plugged. 100ms seems OK - */ - err = readl_poll_timeout(priv->ioaddr + EMAC_BASIC_CTL1, v, - !(v & 0x01), 100, 100000); - - if (err) { + if (err) dev_err(priv->device, "EMAC reset timeout\n"); - return err; - } - return 0; + + return err; } /* Search in mdio-mux node for internal PHY node and get its clk/reset */ @@ -1217,10 +1232,6 @@ static int sun8i_dwmac_probe(struct platform_device *pdev) dev_err(&pdev->dev, "Failed to register mux\n"); goto dwmac_mux; } - } else { - ret = sun8i_dwmac_reset(priv); - if (ret) - goto dwmac_remove; } pm_runtime_put(&pdev->dev); -- 2.50.1 (Apple Git-155)