This is a bug of controller found recently which makes the default values of L1PwrOnSc and L1PwrOnVal unreliable when enabling ASPM. We could work around this by reading L1 substate control 2 register and then write back the value again. Signed-off-by: Shawn Lin <shawn.lin@xxxxxxxxxxxxxx> --- drivers/pci/host/pcie-rockchip.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c index 1671d02..ed782b51 100644 --- a/drivers/pci/host/pcie-rockchip.c +++ b/drivers/pci/host/pcie-rockchip.c @@ -40,6 +40,7 @@ #define PCIE_CLIENT_BASE 0x0 #define PCIE_RC_CONFIG_NORMAL_BASE 0x800000 #define PCIE_RC_CONFIG_BASE 0xa00000 +#define PCIE_RC_CONFIG_L1_SUBSTATE_CTRL2 0x90c #define PCIE_RC_CONFIG_LCSR 0xd0 #define PCIE_RC_CONFIG_LCSR_LBMIE BIT(10) #define PCIE_RC_CONFIG_LCSR_LABIE BIT(11) @@ -477,6 +478,18 @@ static int rockchip_pcie_init_port(struct rockchip_pcie_port *port) return err; } + /* + * We need to read/write PCIE_RC_CONFIG_L1_SUBSTATE_CTRL2 + * before enabling ASPM. Otherwise L1PwrOnSc and L1PwrOnVal isn't + * reliable which makes the controller in broken state when + * enabling ASPM. This is a controller's bug we need to work + * around. + */ + status = pcie_read(port, PCIE_RC_CONFIG_BASE + + PCIE_RC_CONFIG_L1_SUBSTATE_CTRL2); + pcie_write(port, status, PCIE_RC_CONFIG_BASE + + PCIE_RC_CONFIG_L1_SUBSTATE_CTRL2); + /* Enable Gen1 training */ pcie_write(port, HIWORD_UPDATE(PCIE_CLIENT_LINK_TRAIN_ENABLE, -- 2.3.7 -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
