From: Kishon Vijay Abraham I kishon@ti.com
According to errata i802, DCRC error interrupts (MMCHS_STAT[21] DCRC=0x1) can occur during the tuning procedure.
The DCRC interrupt, occurs when the last tuning block fails (the last ratio tested). The delay from CRC check until the interrupt is asserted is bigger than the delay until assertion of the tuning end flag. Assertion of tuning end flag is what masks the interrupts. Because of this race, an erroneous DCRC interrupt occurs.
The suggested workaround is to disable DCRC interrupts during the tuning procedure which is implemented here.
Signed-off-by: Kishon Vijay Abraham I kishon@ti.com Signed-off-by: Jean-Jacques Hiblot jjhiblot@ti.com ---
arch/arm/include/asm/omap_mmc.h | 4 ++++ drivers/mmc/omap_hsmmc.c | 26 ++++++++++++++++++++++---- 2 files changed, 26 insertions(+), 4 deletions(-)
diff --git a/arch/arm/include/asm/omap_mmc.h b/arch/arm/include/asm/omap_mmc.h index 0293281..0893844 100644 --- a/arch/arm/include/asm/omap_mmc.h +++ b/arch/arm/include/asm/omap_mmc.h @@ -219,6 +219,10 @@ struct omap_hsmmc_plat { #define mmc_reg_out(addr, mask, val)\ writel((readl(addr) & (~(mask))) | ((val) & (mask)), (addr))
+#define INT_EN_MASK (IE_BADA | IE_CERR | IE_DEB | IE_DCRC |\ + IE_DTO | IE_CIE | IE_CEB | IE_CCRC | IE_ADMAE | IE_CTO |\ + IE_BRR | IE_BWR | IE_TC | IE_CC) + int omap_mmc_init(int dev_index, uint host_caps_mask, uint f_max, int cd_gpio, int wp_gpio);
diff --git a/drivers/mmc/omap_hsmmc.c b/drivers/mmc/omap_hsmmc.c index fb29a08..3cfd062 100644 --- a/drivers/mmc/omap_hsmmc.c +++ b/drivers/mmc/omap_hsmmc.c @@ -476,6 +476,25 @@ tuning_error: #endif #endif
+static void mmc_enable_irq(struct mmc *mmc, struct mmc_cmd *cmd) +{ + struct omap_hsmmc_data *priv = omap_hsmmc_get_data(mmc); + struct hsmmc *mmc_base = priv->base_addr; + u32 irq_mask = INT_EN_MASK; + + /* + * TODO: Errata i802 indicates only DCRC interrupts can occur during + * tuning procedure and DCRC should be disabled. But see occurences + * of DEB, CIE, CEB, CCRC interupts during tuning procedure. These + * interrupts occur along with BRR, so the data is actually in the + * buffer. It has to be debugged why these interrutps occur + */ + if (cmd && mmc_is_tuning_cmd(cmd->cmdidx)) + irq_mask &= ~(IE_DEB | IE_DCRC | IE_CIE | IE_CEB | IE_CCRC); + + writel(irq_mask, &mmc_base->ie); +} + static int omap_hsmmc_init_setup(struct mmc *mmc) { struct omap_hsmmc_data *priv = omap_hsmmc_get_data(mmc); @@ -542,10 +561,7 @@ static int omap_hsmmc_init_setup(struct mmc *mmc)
writel(readl(&mmc_base->hctl) | SDBP_PWRON, &mmc_base->hctl);
- writel(IE_BADA | IE_CERR | IE_DEB | IE_DCRC | IE_DTO | IE_CIE | - IE_CEB | IE_CCRC | IE_ADMAE | IE_CTO | IE_BRR | IE_BWR | IE_TC | - IE_CC, &mmc_base->ie); - + mmc_enable_irq(mmc, NULL); mmc_init_stream(mmc_base);
return 0; @@ -810,6 +826,8 @@ static int omap_hsmmc_send_cmd(struct udevice *dev, struct mmc_cmd *cmd, #endif }
+ mmc_enable_irq(mmc, cmd); + writel(cmd->cmdarg, &mmc_base->arg); udelay(20); /* To fix "No status update" error on eMMC */ writel((cmd->cmdidx << 24) | flags, &mmc_base->cmd);