Hi,

On Mon, Aug 22, 2011 at 11:10:21PM -0600, Paul Walmsley wrote:
> 
> Convert the omap-serial hardirq handler to a threaded IRQ handler. Without 
> this patch, OMAP boards which use the on-board OMAP UARTs and the 
> omap-serial driver will not boot to userspace after commit 
> f3637a5f2e2eb391ff5757bc83fb5de8f9726464 ("irq: Always set IRQF_ONESHOT if 
> no primary handler is specified").  Enabling CONFIG_DEBUG_SHIRQ reveals 
> 'IRQ handler type mismatch' errors:
> 
> IRQ handler type mismatch for IRQ 74
> current handler: serial idle
> [<c001b124>] (unwind_backtrace+0x0/0xf0) from [<c009c900>] (__setup_irq+0x42c/0x464)
> [<c009c900>] (__setup_irq+0x42c/0x464) from [<c009ca08>] (request_threaded_irq+0xd0/0x148)
> [<c009ca08>] (request_threaded_irq+0xd0/0x148) from [<c029d398>] (serial_omap_startup+0x30/0x2dc)
> [<c029d398>] (serial_omap_startup+0x30/0x2dc) from [<c0295a38>] (uart_startup+0x5c/0x1ac)
> 
> (etc.)
> 
> It turns out that the omap-serial code used one threaded IRQ
> handler[1][2] and one non-threaded IRQ handler[3] that shared the same
> IRQ.  During the 3.1-rc series, a patch was merged[4] that caused
> IRQF_ONESHOT to be set on the threaded handler, but the non-threaded
> handler did not have IRQF_ONESHOT set.  Since interrupt handlers
> that share the same IRQ must also share the presence or absence of
> IRQF_ONESHOT[5], this new commit caused a mismatch that prevented the
> non-threaded IRQ from registering.
> 
> Fix by converting the non-threaded IRQ handler in omap-serial.c to a
> threaded IRQ handler.  Ideally we would not have to make this type of
> change during the -rc series, but the commit that caused this behavior
> was itself merged between v3.1-rc2 and v3.1-rc3.  In the long term, it
> would be good to get rid of the shared IRQ handler hack in
> arch/arm/mach-omap2/serial.c.
> 
> This change has been boot-tested on OMAP3530 BeagleBoard and
> OMAP4430ES2 PandaBoard, and static suspend has been lightly tested on
> the BeagleBoard.
> 
> Pantelis Antoniou <pantelis.antoniou@gmail.com> originally reported
> the boot failure.
> 
> 1. arch/arm/mach-omap2/serial.c line 550, as of Linux v3.1-rc3.  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/arm/mach-omap2/serial.c;h=466fc722fa0f39f03b8d93cf84e4dae4f57fd029;hb=fcb8ce5cfe30ca9ca5c9a79cdfe26d1993e65e0c#l550
> 
> 2. arch/arm/mach-omap2/serial.c line 563, as of Linux v3.1-rc3.  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/arm/mach-omap2/serial.c;h=466fc722fa0f39f03b8d93cf84e4dae4f57fd029;hb=fcb8ce5cfe30ca9ca5c9a79cdfe26d1993e65e0c#l563
> 
> 3. drivers/tty/serial/omap-serial.c line 464, as of Linux v3.1-rc3.  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=drivers/tty/serial/omap-serial.c;h=c37df8d0fa2819261dffccc5bc4d0180b9531f49;hb=fcb8ce5cfe30ca9ca5c9a79cdfe26d1993e65e0c#l464
> 
> 4. Commit f3637a5f2e2eb391ff5757bc83fb5de8f9726464 ("irq: Always set
>    IRQF_ONESHOT if no primary handler is specified")
> 
> 5. Gleixner, Thomas. _[patch 2/5] genirq: Allow shared oneshot
>    interrupts_.  http://lkml.org/lkml/2011/2/23/511
> 
> Signed-off-by: Paul Walmsley <paul@pwsan.com>
> Cc: Pantelis Antoniou <pantelis.antoniou@gmail.com>
> Cc: Govindraj.R <govindraj.raja@ti.com>
> Cc: Kevin Hilman <khilman@ti.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> ---
>  drivers/tty/serial/omap-serial.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/tty/serial/omap-serial.c b/drivers/tty/serial/omap-serial.c
> index c37df8d..e83a769 100644
> --- a/drivers/tty/serial/omap-serial.c
> +++ b/drivers/tty/serial/omap-serial.c
> @@ -461,8 +461,8 @@ static int serial_omap_startup(struct uart_port *port)
>  	/*
>  	 * Allocate the IRQ
>  	 */
> -	retval = request_irq(up->port.irq, serial_omap_irq, up->port.irqflags,
> -				up->name, up);
> +	retval = request_threaded_irq(up->port.irq, NULL, serial_omap_irq,
> +				      up->port.irqflags, up->name, up);

if you're not running on a slow bus, you should use top half to check if
IRQs are really from this device.

-- 
balbi