Re: [PATCH]: Atmel Serial Console interrupt handler splitup

[Remy Bohmer <linux@bohmer.net>]

Hi,

2010/10/29 Aras Vaichas <arasv@magellan-technology.com>:
>> >                On AT91 David Brownell noticed several months
>> > ago that the DBGU can miss some characters on NO_HZ. I would expect
>> > that this would be better now due to the shorter interrupt handler,
>> > although it was not my goal to solve it with these patches.
>> The current (kernel.org git) AT91 timer handler incorporates a bugfix
>> in that area, which seemed to resolve most of those problems.
>> - Dave
>
> Hello,
>
> I'm sorry to bother you two off-list, but I'm trying to avoid too much
> noise on the lists.

Please, always at least CC to the mailinglist as well.

> My issue is related to these two posts (and the re: above):
>
> http://www.mail-archive.com/linux-rt-users@vger.kernel.org/msg02115.html
> http://permalink.gmane.org/gmane.linux.ports.arm.kernel/94652
>
> If I change the Atmel serial interrupt to IRQF_NODELAY, then *most* of
> the overruns go away and DBGU works as it should.
>
> --- A/drivers/serial/atmel_serial.c   2010-08-03 03:27:18.000000000 +1000
> +++ B/drivers/serial/atmel_serial.c 2010-10-29 10:24:54.000000000 +1100
> @@ -808,7 +818,7 @@
>        /*
>         * Allocate the IRQ
>         */
> -       retval = request_irq(port->irq, atmel_interrupt, IRQF_SHARED,
> +       retval = request_irq(port->irq, atmel_interrupt, IRQF_NODELAY,
>                        tty ? tty->name : "atmel_serial", port);
>        if (retval) {
>                printk("atmel_serial: atmel_startup - Can't get irq\n");
>
>
> But shouldn't the DBGU interrupt already be running in NODELAY context
> according to Remy's post?  I understand that NODELAY has many other
> complications, this is just a test to get an idea of which direction
> to move in.

On preempt-RT this driver always has NODELAY set by default. AFAIK, it
is not in mainline.

Notice that on recent kernels also the request_threaded_irq()
mechanism can be used, which is much cleaner compared to the NODELAY
change.
So, I guess the NODELAY change will never hit mainline since this
better solution exist.

Kind regards,

Remy
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html