Re: [PATCH] Clear previous interrupts after fifo is disabled

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Tue, Feb 28, 2012 at 10:35:30AM +0900, Chanho Min wrote:
> On Mon, Feb 27, 2012 at 8:02 PM, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> > On Mon, Feb 27, 2012 at 10:48:58AM +0000, Russell King - ARM Linux wrote:
> >> I'd much prefer to only clear those interrupts which actually need to be
> >> cleared at this point.  So, I'd suggest this approach instead:
> >
> > Thinking about this a little more, we definitely want to mask and clear
> > interrupts at probe time as well:
> >
> I'm not satisfied with this completely. RIS has some pending
> interrupts even if interrupts are masked/disabled in IMSC. If your
> patch is applied, interrupt can be pended as bellows and RXFE of the
> flag register is set as well.

RXFE _will_ be set.  Think about it - RXFE means Receive Fifo Empty.
If the receive fifo is empty, it _will_ be set.

And RIS is the _Raw_ interrupt status.  That's the status _before_ the
mask is acted upon.

> writew(UART011_OEIS | UART011_BEIS | UART011_PEIS | UART011_FEIS |
>        UART011_RTIS | UART011_RXIS, uap->port.membase + UART011_ICR);
> ...
> Interrupt is occured and pended in RIS

But it won't be delivered because the mask register is zero.

> ..
> writew(uap->im, uap->port.membase + UART011_IMSC);
> 
> Root cause is that Rx interrupt set but Rx fifo is empty. If we just
> remove the sentence for clearing LCRH, nothing happens and interrupt
> handler don't this misbehave.

No.