diff options
author | Imre Deak <imre.deak@nokia.com> | 2006-06-26 16:16:00 -0700 |
---|---|---|
committer | Tony Lindgren <tony@atomide.com> | 2006-06-26 16:16:00 -0700 |
commit | ea6dedd7fbd0f760ebf37eb0bcc8c64856475a13 (patch) | |
tree | 01d4dc723a6fa3fb567dc2d0ce497e69af4d7721 /arch/arm/mach-omap1 | |
parent | 7876284178b386ab6e97f96733ccbc19eda743a3 (diff) | |
download | kernel_samsung_espresso10-ea6dedd7fbd0f760ebf37eb0bcc8c64856475a13.zip kernel_samsung_espresso10-ea6dedd7fbd0f760ebf37eb0bcc8c64856475a13.tar.gz kernel_samsung_espresso10-ea6dedd7fbd0f760ebf37eb0bcc8c64856475a13.tar.bz2 |
ARM: OMAP: GPIO IRQ lazy IRQ disable fix
- The current OMAP GPIO IRQ framework doesn't use the do_edge_IRQ,
do_level_IRQ handlers, but instead calls do_simple_IRQ. This
doesn't handle disabled interrupts properly, so drivers will
still get interrupts after calling disable_irq. The patch solves
this by respecting the irq_desc.disable_depth and irq_desc.running
counters. When one of these is non-zero the handler is not called,
the interrupt is masked and marked as pending. The pending interrupt
will be serviced when the running handler returns. This is according
to the same semantics as the standard do_edge_IRQ and do_level_IRQ
handlers have, so one day we should use them instead of do_simple_IRQ.
- Process only interrupts that are not masked. The ISR may contain
pending interrupts that are masked these shouldn't be processed.
- Move the bank IRQ unmasking out of the IRQ dispatch loop. If there
are further iterations we shouldn't unmask it if there are level
triggered interrupts pending.
Signed-off-by: Imre Deak <imre.deak@nokia.com>
Signed-off-by: Juha Yrjola <juha.yrjola@nokia.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Diffstat (limited to 'arch/arm/mach-omap1')
0 files changed, 0 insertions, 0 deletions