From mboxrd@z Thu Jan 1 00:00:00 1970 From: Scott Wood Date: Thu, 21 Mar 2013 17:28:30 -0500 Subject: [U-Boot] [PATCH] omap3_beagle: Enable CONFIG_SYS_NS16550_BROKEN_TEMT In-Reply-To: <20130321222100.GK26945@bill-the-cat> (from trini@ti.com on Thu Mar 21 17:21:00 2013) Message-ID: <1363904910.31522.27@snotra> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On 03/21/2013 05:21:00 PM, Tom Rini wrote: > On Thu, Mar 21, 2013 at 08:03:59PM +0100, Manfred Huber wrote: > > From: Manfred Huber > > > > Beagleboard UART (ns16550) doesn't set the Transmitter Empty (TEMT) > > Bit in SPL. The serial port behaves differently based on the stage of U-Boot that is running? Or is it that the bit doesn't get set until the port has been properly initialized? Couldn't that happen in a case where SPL isn't used at all? > > Only Transmitter Hold Register Empty (THRE) Bit is set. > > This makes SPL to hang while waiting for TEMT. Adding the > > CONFIG_SYS_NS16550_BROKEN_TEMT config option and waiting for THRE > > avoid this issue. > > > > Signed-off-by: Manfred Huber > > --- > > drivers/serial/ns16550.c | 5 ++++- > > include/configs/omap3_beagle.h | 3 +++ > > 2 files changed, 7 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/serial/ns16550.c b/drivers/serial/ns16550.c > > index b2da8b3..6379bcc 100644 > > --- a/drivers/serial/ns16550.c > > +++ b/drivers/serial/ns16550.c > > @@ -36,7 +36,10 @@ > > > > void NS16550_init(NS16550_t com_port, int baud_divisor) > > { > > -#if (!defined(CONFIG_SYS_NS16550_BROKEN_TEMT)) > > +#if defined(CONFIG_SPL_BUILD) && > defined(CONFIG_SYS_NS16550_BROKEN_TEMT) > > + while (!(serial_in(&com_port->lsr) & UART_LSR_THRE)) > > + ; > > +#else > > while (!(serial_in(&com_port->lsr) & UART_LSR_TEMT)) > > ; > > #endif > > Scott, do you still have access to the failing systems that made us > introduce this change to start with? It was an intermittent failure seen on a development tree, so not really. You could try testing it by printing something immediately before calling NS16550_init(), either in a situation where you know the serial port is already configured (e.g. by SPL) or by calling NS16550_init() twice. > Could we perhaps go with the THRE test instead in all cases? Thanks! Wouldn't that still allow the last character to possibly be corrupted? -Scott