ENGR00319415 pcie: random link down issue after warm-rst

There are about 0.02% percentage on some imx6q/dl/solo
hw boards, random pcie link down when warm-reset is used.
Make sure to clear the ref_ssp_en bit16 of gpr1 before
warm-rst, and set ref_ssp_en after the pcie clks are
stable to workaround it.

rootcause:
* gpr regisers wouldn't be reset by warm-rst, while the
ref_ssp_en is required to be reset by pcie.
(work-around in u-boot)
* ref_ssp_en should be set after pcie clks are stable.
(work-around in kernel)

Signed-off-by: Richard Zhu <r65037@freescale.com>
(cherry picked from commit 5cc825b12c6b86a22f1a6a0535b52cf3ee142e77)
Signed-off-by: Peng Fan <Peng.Fan@freescale.com>
(cherry picked from commit 6193cf4e3384a59e29546d13a67657f7faeafc9e)

1 files changed, 18 insertions, 0 deletions

diff --git a/arch/arm/cpu/armv7/mx6/soc.c b/arch/arm/cpu/armv7/mx6/soc.c
index a71de6e..ba8853e 100644
--- a/arch/arm/cpu/armv7/mx6/soc.c
+++ b/arch/arm/cpu/armv7/mx6/soc.c
@@ -413,6 +413,24 @@ static void imx_set_pcie_phy_power_down(void)
 
 int arch_cpu_init(void)
 {
+#ifndef CONFIG_MX6SX
+	/* this bit is not used by imx6sx anymore */
+	u32 val;
+
+	/*
+	 * There are about 0.02% percentage, random pcie link down
+	 * when warm-reset is used.
+	 * clear the ref_ssp_en bit16 of gpr1 to workaround it.
+	 * then warm-reset imx6q/dl/solo again.
+	 */
+	val = readl(IOMUXC_BASE_ADDR + 0x4);
+	if (val & (0x1 << 16)) {
+		val &= ~(0x1 << 16);
+		writel(val, IOMUXC_BASE_ADDR + 0x4);
+		reset_cpu(0);
+	}
+#endif
+
 	init_aips();
 
 	/* Need to clear MMDC_CHx_MASK to make warm reset work. */