[PATCH 1/9] omap3/4: uart: fix full-fifo write abort

[Tony Lindgren]

* Shilimkar, Santosh <santosh.shilimkar at ti.com> [100223 08:14]:
> Tony,
> > -----Original Message-----
> > From: Tony Lindgren [mailto:tony at atomide.com]
> > Sent: Tuesday, February 23, 2010 3:10 AM
> > To: Shilimkar, Santosh
> > Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Woodruff, Richard; Ghorai,
> > Sukumar
> > Subject: Re: [PATCH 1/9] omap3/4: uart: fix full-fifo write abort
> > 
> > * Shilimkar, Santosh <santosh.shilimkar at ti.com> [100218 21:22]:
> > > Bye the way just to add bit more clarity, this patch addresses TX
> > > hardware restriction in the new UART IP used on omap3630 and omap4430.
> > > First part of the fix for RX is already in mainline,
> > > Commit: "ce13d4716a276f4331d78ba28a5093a63822ab95"
> > >
> > > More details on this thread are here.
> > > http://www.mail-archive.com/linux-omap@vger.kernel.org/msg19447.html
> > 
> > Thanks, I've updated the comments for this patch with the link above
> > and added the whole series into omap for-next.
> > 
> Please drop this patch since it has introduces regression on interrupt latency
> while traces are enabled. Since 8250 driver calls serial_out() function is with
> interrupt disabled ( spin_lock_irqsave), the interrupt latency at times is 
> as high as in few milliseconds because of error case.
> 
> I shall work towards this issue but for now we should drop this since the side effect
> of this change is known

OK, dropping it. We can merge something better as a fix during the -rc cycle
once you have it working.

Regards,

Tony

 
> > 
> > > > -----Original Message-----
> > > > From: Shilimkar, Santosh
> > > > Sent: Thursday, February 18, 2010 2:29 PM
> > > > To: tony at atomide.com
> > > > Cc: linux-omap at vger.kernel.org; linux-arm-kernel at lists.infradead.org; Shilimkar, Santosh;
> > Woodruff,
> > > > Richard; Ghorai, Sukumar
> > > > Subject: [PATCH 1/9] omap3/4: uart: fix full-fifo write abort
> > > >
> > > > This patch is addition to the already merged commit on non-empty
> > > > uart fifo read abort. "ce13d4716a276f4331d78ba28a5093a63822ab95"
> > > >
> > > > OMAP3630 and OMAP4430 UART IP blocks have a restriction on TX FIFO
> > > > too. If you try to write to the tx fifo when it is full, the system aborts.
> > > >
> > > > This can be easily reproducible by not suppressing interconnect errors or
> > > > long duration testing where continuous prints over console from multiple
> > > > threads. This patch is addressing the issue by ensuring that write is
> > > > not issued while fifo is full. A timeout is added to avoid any hang
> > > > on fifo-full for 10 mS which is unlikely case.
> > > >
> > > > Patch is validated on OMAP3630 and OMAP4 SDP.
> > > >
> > > > V2 version removed the additional 1 uS on every TX as per
> > > > Tony's suggestion
> > > >
> > > > Signed-off-by: Woodruff Richard <r-woodruff2 at ti.com>
> > > > Signed-off-by: Santosh Shilimkar <santosh.shilimkar at ti.com>
> > > > CC: Ghorai Sukumar <s-ghorai at ti.com>
> > > > ---
> > > >  arch/arm/mach-omap2/serial.c |   31 ++++++++++++++++++++++++++++---
> > > >  1 files changed, 28 insertions(+), 3 deletions(-)
> > > >
> > > > diff --git a/arch/arm/mach-omap2/serial.c b/arch/arm/mach-omap2/serial.c
> > > > index 5f3035e..b79bc89 100644
> > > > --- a/arch/arm/mach-omap2/serial.c
> > > > +++ b/arch/arm/mach-omap2/serial.c
> > > > @@ -23,6 +23,7 @@
> > > >  #include <linux/serial_reg.h>
> > > >  #include <linux/clk.h>
> > > >  #include <linux/io.h>
> > > > +#include <linux/delay.h>
> > > >
> > > >  #include <plat/common.h>
> > > >  #include <plat/board.h>
> > > > @@ -160,6 +161,13 @@ static inline unsigned int serial_read_reg(struct plat_serial8250_port *up,
> > > >  	return (unsigned int)__raw_readb(up->membase + offset);
> > > >  }
> > > >
> > > > +static inline void __serial_write_reg(struct uart_port *up, int offset,
> > > > +		int value)
> > > > +{
> > > > +	offset <<= up->regshift;
> > > > +	__raw_writeb(value, up->membase + offset);
> > > > +}
> > > > +
> > > >  static inline void serial_write_reg(struct plat_serial8250_port *p, int offset,
> > > >  				    int value)
> > > >  {
> > > > @@ -620,6 +628,20 @@ static unsigned int serial_in_override(struct uart_port *up, int offset)
> > > >  	return __serial_read_reg(up, offset);
> > > >  }
> > > >
> > > > +static void serial_out_override(struct uart_port *up, int offset, int value)
> > > > +{
> > > > +	unsigned int status, tmout = 10000;
> > > > +
> > > > +	status = __serial_read_reg(up, UART_LSR);
> > > > +	while (!(status & UART_LSR_THRE)) {
> > > > +		/* Wait up to 10ms for the character(s) to be sent. */
> > > > +		if (--tmout == 0)
> > > > +			break;
> > > > +		udelay(1);
> > > > +		status = __serial_read_reg(up, UART_LSR);
> > > > +	}
> > > > +	__serial_write_reg(up, offset, value);
> > > > +}
> > > >  void __init omap_serial_early_init(void)
> > > >  {
> > > >  	int i;
> > > > @@ -721,11 +743,14 @@ void __init omap_serial_init_port(int port)
> > > >  	 * omap3xxx: Never read empty UART fifo on UARTs
> > > >  	 * with IP rev >=0x52
> > > >  	 */
> > > > -	if (cpu_is_omap44xx())
> > > > +	if (cpu_is_omap44xx()) {
> > > >  		uart->p->serial_in = serial_in_override;
> > > > -	else if ((serial_read_reg(uart->p, UART_OMAP_MVER) & 0xFF)
> > > > -			>= UART_OMAP_NO_EMPTY_FIFO_READ_IP_REV)
> > > > +		uart->p->serial_out = serial_out_override;
> > > > +	} else if ((serial_read_reg(uart->p, UART_OMAP_MVER) & 0xFF)
> > > > +			>= UART_OMAP_NO_EMPTY_FIFO_READ_IP_REV) {
> > > >  		uart->p->serial_in = serial_in_override;
> > > > +		uart->p->serial_out = serial_out_override;
> > > > +	}
> > > >  }
> > > >
> > > >  /**
> > > > --
> > > > 1.6.0.4
> > >
> 
> Regards,
> Santosh