2009-09-18 09:04:50

by Sonic Zhang

[permalink] [raw]
Subject: [PATCH][bfin-sport-uart] Walk around RX underflow hardware issue in 32-bit SPORT RX reading

Disable local interrupts in 32-bit reading from SPORT RX MMR to walk
around the RX underflow hardware issue when this reading is interrupted
by the other IRQ.


Signed-off-by: Sonic Zhang <[email protected]>
---
drivers/serial/bfin_sport_uart.h | 15 ++++++++++++++-
1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/drivers/serial/bfin_sport_uart.h b/drivers/serial/bfin_sport_uart.h
index 671d41c..46e793e 100644
--- a/drivers/serial/bfin_sport_uart.h
+++ b/drivers/serial/bfin_sport_uart.h
@@ -43,7 +43,20 @@
#define SPORT_GET_TFSDIV(sport) bfin_read16(((sport)->port.membase + OFFSET_TFSDIV))
#define SPORT_GET_TX(sport) bfin_read16(((sport)->port.membase + OFFSET_TX))
#define SPORT_GET_RX(sport) bfin_read16(((sport)->port.membase + OFFSET_RX))
-#define SPORT_GET_RX32(sport) bfin_read32(((sport)->port.membase + OFFSET_RX))
+/*
+ * Disable local interrupt to walk around the fake RX underflow error
+ * when do 32-bit reading from RX fifois interrupted by other interrupt.
+ * This is a possible hardware anomaly. After it is logged officially,
+ * the official anomaly id should be listed bellow.
+ */
+#define SPORT_GET_RX32(sport) \
+({ \
+ unsigned int __ret; \
+ local_irq_disable(); \
+ __ret = bfin_read32(((sport)->port.membase + OFFSET_RX)); \
+ local_irq_enable(); \
+ __ret; \
+})
#define SPORT_GET_RCR1(sport) bfin_read16(((sport)->port.membase + OFFSET_RCR1))
#define SPORT_GET_RCR2(sport) bfin_read16(((sport)->port.membase + OFFSET_RCR2))
#define SPORT_GET_RCLKDIV(sport) bfin_read16(((sport)->port.membase + OFFSET_RCLKDIV))
--
1.6.0



2009-09-18 15:14:57

by Mike Frysinger

[permalink] [raw]
Subject: Re: [PATCH][bfin-sport-uart] Walk around RX underflow hardware issue in 32-bit SPORT RX reading

On Fri, Sep 18, 2009 at 05:05, sonic zhang wrote:
> Disable local interrupts in 32-bit reading from SPORT RX MMR to walk
> around the RX underflow hardware issue when this reading is interrupted
> by the other IRQ.

ive cleaned up the code in svn, so if you want to grab the updated
version for sending out.
Signed-off-by: Mike Frysinger <[email protected]>
-mike