Re: Serial driver problems with 2.1.120-pre3

Richard Gooch (rgooch@atnf.csiro.au)
Wed, 2 Sep 1998 22:48:33 +1000


Etienne Lorrain writes:
> Richard Gooch <rgooch@atnf.csiro.au> wrote:
> > Hi, all. I've just tried 2.1.120-pre3 and I've noticed an unpleasant
> > interaction with setserial v2.14. If I do:
> >
> > # setserial /dev/ttyS0 auto_irq skip_test autoconfig
> > # setserial /dev/ttyS1 auto_irq skip_test autoconfig
> > # setserial -bg /dev/ttyS0 /dev/ttyS1
> > /dev/ttyS0 at 0x03f8 (irq = 0) is a 16550A
> > /dev/ttyS1 at 0x02f8 (irq = 0) is a 16550A
>
> I do not know if it is related to 2.1.120-pre3, but this
> can appear if the two serial ports are on the same IRQ, or
> if another board maintains the IRQ line active.
> Also, you have a two processor machine, the two "setserial"
> may be done in parallel - not nice for IRQ probing...
>
> Why did you add the "skip_test" option ?

Erm, because I'm using rc.serial from setserial v2.01.

> What happens if you compile the kernel with "SERIAL_DETECT_IRQ"
> (then the driver initialise all serial ports - and clear IRQ lines -
> *before* autodetecting the IRQs, and does that sequencially for
> each lines).
>
> Then you can remove the setserial lines from your startup scripts,
> probing for IRQ is basically unsafe but it is better to do that
> at kernel startup than when a lot of tasks are running in 'init'.
> The safest is always to do "setserial /dev/ttyS0 irq 4".

Well, as I recall, a long time ago we had auto IRQ probing in the
serial driver, but Ted disabled it because he felt it was unneeded in
the kernel and it was somehow "better" to use setserial (I don't
recall any more detail: it was years ago). I think it was targetted
for eventual removal.
So, re-enabling auto IRQ detection in the kernel is probably not what
Ted wants. Given that, I'd rather get setserial working
again. Especially considering that the default kernel configuration is
no auto IRQ probing. It would seem wise to get the default case
working :-)

Ted: can you enlighten us further?

Regards,

Richard....

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html