From: linux@arm.linux.org.uk (Russell King - ARM Linux) Date: Thu, 29 Oct 2009 19:12:21 +0000 Subject: [PATCH] iop13xx: workaround errata that causes uart interrupts to be missed In-Reply-To: <1256843340.28922.14.camel@dwillia2-linux.ch.intel.com> References: <1256843340.28922.14.camel@dwillia2-linux.ch.intel.com> Message-ID: <20091029191221.GD11460@n2100.arm.linux.org.uk> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org You need to cc Alan Cox and lkml on this. On Thu, Oct 29, 2009 at 12:09:00PM -0700, Dan Williams wrote: > Reading the uart iir register sometimes results in the interrupt being > cleared without being reported. Use the the system interrupt pending > register as a fallback. > > This converts locations in the code that only care about the interrupt > pending bit to use the iir_int_status() macro. > > Signed-off-by: Dan Williams > --- > This patch has been sitting in the 'misc' branch of the xscaleiop tree > in an un-mergeable form for quite a while [1]. I finally got fed up > with forward porting it and would like to try to get this workaround > upstream. There appears to be no maintainer for 8250.c, so not sure who > should merge this. I can append it to the end of my next pull request > for the iop queue...? > > Thanks, > Dan > > [1]: http://git.kernel.org/?p=linux/kernel/git/djbw/xscaleiop.git;a=commitdiff;h=a7d124f5 > > drivers/serial/8250.c | 36 +++++++++++++++++++++++++++++++++--- > 1 files changed, 33 insertions(+), 3 deletions(-) > > diff --git a/drivers/serial/8250.c b/drivers/serial/8250.c > index b1ae774..c905860 100644 > --- a/drivers/serial/8250.c > +++ b/drivers/serial/8250.c > @@ -547,6 +547,36 @@ serial_out_sync(struct uart_8250_port *up, int offset, int value) > (up->port.serial_in(&(up)->port, (offset))) > #define serial_out(up, offset, value) \ > (up->port.serial_out(&(up)->port, (offset), (value))) > + > +#ifndef CONFIG_ARCH_IOP13XX > +#define iir_int_status(up) serial_in((up), UART_IIR) > +#else > +/* A silicon issue on IOP13XX causes IIR reads to return UART_IIR_NO_INT > + * even if an interrupt is pending. Use the system interrupt pending > + * register to check for servicing as a workaround. > + */ > +#include > +static unsigned int iir_int_status(struct uart_8250_port *up) > +{ > + unsigned long membase = (unsigned long) up->port.membase; > + unsigned long ipnd, iir; > + > + ipnd = read_intpnd_1(); > + iir = UART_IIR_NO_INT; > + if (membase == IOP13XX_UART0_VIRT) { > + if (ipnd & (1 << (IRQ_IOP13XX_UART0 - 32))) > + iir &= ~UART_IIR_NO_INT; > + } else if (membase == IOP13XX_UART1_VIRT) { > + if (ipnd & (1 << (IRQ_IOP13XX_UART1 - 32))) > + iir &= ~UART_IIR_NO_INT; > + } else > + BUG(); > + /* clear the interupt via the LSR */ > + mem_serial_in(&up->port, UART_LSR); > + return iir; > +} > +#endif > + > /* > * We used to support using pause I/O for certain machines. We > * haven't supported this for a while, but just in case it's badly > @@ -1572,7 +1602,7 @@ static irqreturn_t serial8250_interrupt(int irq, void *dev_id) > > up = list_entry(l, struct uart_8250_port, list); > > - iir = serial_in(up, UART_IIR); > + iir = iir_int_status(up); > if (!(iir & UART_IIR_NO_INT)) { > serial8250_handle_port(up); > > @@ -1732,7 +1762,7 @@ static void serial8250_timeout(unsigned long data) > struct uart_8250_port *up = (struct uart_8250_port *)data; > unsigned int iir; > > - iir = serial_in(up, UART_IIR); > + iir = iir_int_status(up); > if (!(iir & UART_IIR_NO_INT)) > serial8250_handle_port(up); > mod_timer(&up->timer, jiffies + poll_timeout(up->port.timeout)); > @@ -1753,7 +1783,7 @@ static void serial8250_backup_timeout(unsigned long data) > serial_out(up, UART_IER, 0); > } > > - iir = serial_in(up, UART_IIR); > + iir = iir_int_status(up); > > /* > * This should be a safe test for anyone who doesn't trust the > >