[ This is the second version of these patches - they now check the
SIH_CTRL.COR bit to determine whether to read or write to clear ]
At boot, twl_init_irq() tries to mask off all TWL4030 secondary
interrupts and clear them, since no secondary interrupt handlers are
registered yet (that happens later in the boot).
However, this code is flawed. MADC, BCI, keypad, and GPIO ISRs are
either not cleared or are incompletely cleared. Some BCI register
addresses are nonsensical. And some code attempts to clear ISRs
before masking IMRs, which is race-prone (an interrupt could occur
after the ISR clear, but before the mask).
These problems can cause large numbers of boot messages similar to
this:
TWL4030 module irq 368 is disabled but can't be masked!
and I2C transmit overflows, both observed during boot on the
BeagleBoard here. The actual IRQ number sometimes varied depending on
which interrupt was unacknowledged.
This series fixes these issues and does some code cleanup, saving 272
bytes.
Thanks to Philip Balister <philip@xxxxxxxxxxxx>, Koen Kooi
<k.kooi@xxxxxxxxxxxxxxxxxx>, Måns Rullgård <mans@xxxxxxxxx>, Steve
Sakoman <sakoman@xxxxxxxxx> and others on #beagle for help tracing
these problems.
Thanks also to Peter De Schrijver <peter.de-schrijver@xxxxxxxxx> for
help with this second version, allowing several unnecessary I2C
transactions to be elided from twl_init_irq().
Tested on BeagleBoard rev B4 (3530ES2.2).
- Paul
---
size:
text data bss dec hex filename
3076157 142840 104248 3323245 32b56d vmlinux.beagle.orig
3075885 142840 104248 3322973 32b45d vmlinux.beagle
diffstat:
drivers/i2c/chips/twl4030-core.c | 383 ++++++++++++++++++++++--------------
drivers/i2c/chips/twl4030-gpio.c | 48 -----
drivers/i2c/chips/twl4030-pwrirq.c | 15 +
include/linux/i2c/twl4030-gpio.h | 76 +++++++
include/linux/i2c/twl4030-pwrirq.h | 37 +++
5 files changed, 359 insertions(+), 200 deletions(-)
create mode 100644 include/linux/i2c/twl4030-gpio.h
create mode 100644 include/linux/i2c/twl4030-pwrirq.h
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
