From 94d32f99c6aba8b702a9cf78cb328aeb7882b1e5 Mon Sep 17 00:00:00 2001 From: Andy Green Date: Mon, 1 Feb 2010 13:28:54 +0100 Subject: imx-serial-excessive-status-clearing-on-break.patch cc: Sasha Hauer We noticed that we were only able to communicate once with a GSM module per session, if we powered down the module then the serial driver was broken until reset, no traffic or /proc/interrupts activity could be seen any more. Volker noticed it was provoked by a long "break" seen on the wire when the RX was not driven during powerdown, we discovered we could kill the serial driver just by forcing RX low extenally for a little while. I saw that the processing for BREAK in the driver gets too excited and clears down all set bits in USR2, including that a character is pending. This stops all further characters getting processed. The attached one-liner makes the serial driver immortal against BREAK. Signed-off-by: Andy Green Signed-off-by: Volker Ernst Signed-off-by: Sascha Hauer --- drivers/serial/imx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/serial/imx.c') diff --git a/drivers/serial/imx.c b/drivers/serial/imx.c index 18130f11238e..c5b546a98520 100644 --- a/drivers/serial/imx.c +++ b/drivers/serial/imx.c @@ -440,7 +440,7 @@ static irqreturn_t imx_rxint(int irq, void *dev_id) temp = readl(sport->port.membase + USR2); if (temp & USR2_BRCD) { - writel(temp | USR2_BRCD, sport->port.membase + USR2); + writel(USR2_BRCD, sport->port.membase + USR2); if (uart_handle_break(&sport->port)) continue; } -- cgit v1.2.1