mmc: sh_mmcif: don't clear masked interrupts

Masking events on MMCIF means that an occurrence of the masked event
won't raise an interrupt, but the event bit will still be set in the
interrupt status register. If simultaneously a different event occurs
which was enabled, both flags will be set. However, only the unmasked
event bit should be cleared in the status register in such a case.

Clearing also the masked bit can lead to lost interrupts, which indeed
can be observed on the armadillo800eva r8a7740 board with an eMMC chip.
The problem has been introduced by the recent "mmc: sh_mmcif: simplify
IRQ processing" patch. Fix the problem by only clearing enabled interrupts.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Tested-by: Nguyen Viet Dung <nv-dung@jinso.co.jp>
Tested-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Chris Ball <cjb@laptop.org>

diff --git a/drivers/mmc/host/sh_mmcif.c b/drivers/mmc/host/sh_mmcif.c
index 117a1f7747201b71233d991a6d64f103d0dbc3dd..8ef5efad45d0a9d68a879913601c841b32af9f1d 100644 (file)
--- a/drivers/mmc/host/sh_mmcif.c
+++ b/drivers/mmc/host/sh_mmcif.c
@@ -1244,7 +1244,8 @@ static irqreturn_t sh_mmcif_intr(int irq, void *dev_id)
        u32 state;
 
        state = sh_mmcif_readl(host->addr, MMCIF_CE_INT);
-       sh_mmcif_writel(host->addr, MMCIF_CE_INT, ~state);
+       sh_mmcif_writel(host->addr, MMCIF_CE_INT,
+                       ~(state & sh_mmcif_readl(host->addr, MMCIF_CE_INT_MASK)));
        sh_mmcif_bitclr(host, MMCIF_CE_INT_MASK, state & MASK_CLEAN);
 
        if (state & ~MASK_CLEAN)