On Wed, Sep 12, 2012 at 10:44:20AM +0100, Alan Cox wrote:

quoted

Of course, if Alan is OK with this, I'm more than OK too. :-)

It may well be better.

quoted

(But the polling routines would need to clear all interrupts, not
just rx/tx. For example, if the controller indicated some error, and
nobody clears it, then we'll start reentering infinitely.)

For a lot of devices and platforms you'd probably mask them instead ?

If there is no way to clear them, yes, we obviously would want to
mask them before using the port for NMI debugger. Then we'd need
three callbacks:

- mask_all_irqs_but_rx() -- used before we want to start using the port
  for the NMI debugger;
- clear_rx_irq() -- (optional) clears rx IRQ for controllers that need
  it;
- restore_irqs() -- unmasks interrupts that were previously masked.

If we ever encounter a case when just clearing interrupts doesn't work,
we can surely implement the above scheme... It's just so far I don't
see any need to over-design this, but again, it's your call, I told my
opinion on this, but I'll do whatever you guys like more. :-)

quoted

quoted

If you use a clear_irqs callback, you can drop characters if
one arrives between the last character buffer read and calling
clear_irqs.

Only if we call clear_irqs() after reading the characters, but we do
it before. So if new characters are available, we will reenter NMI,
which is OK.

Recursively or not... again you get platform specific magic in places
we don't want.

I really really don't see how this is platform-specific. All we ask the
serial driver is to quiesce its interrupt. Whether we can handle
NMIs/IRQs recursively or not is not serial driver's worry, since its
IRQ handler is not going to fire anyway. The polling routines already
gave us the power to steal/inject the data, so now we're stealing the
interrupt too.

How we use the callback is indeed platform-specific, but so is the
whole KGDB, and that knowledge is hidden there.

For serial driver it's all pretty much clear: lower the interrupt, but
don't turn off rx detection.

Thanks!

Anton.