netconsole and a serial one

netconsole and a serial one

[Jon Smirl]

I have both netconsole and a serial console setup on an embedded
system. This is what I need for the boot process.

After I'm booted I want to turn off printk() output to the serial
console but keep it going to netconsole.
I've been googling for an hour, how do I do this?

I need to turn off the serial output after boot because serial
printk() disables interrupts for up to 1ms which is enough to cause me
to lose a lot of other interrupts.

-- 
Jon Smirl
jonsmirl@gmail.com

Re: netconsole and a serial one

[Jeff Moyer]

"Jon Smirl" <jonsmirl@gmail.com> writes:

> I have both netconsole and a serial console setup on an embedded
> system. This is what I need for the boot process.
>
> After I'm booted I want to turn off printk() output to the serial
> console but keep it going to netconsole.
> I've been googling for an hour, how do I do this?
>
> I need to turn off the serial output after boot because serial
> printk() disables interrupts for up to 1ms which is enough to cause me
> to lose a lot of other interrupts.

There's no way (that I know of) to do this presently.  However, I think
David Howells has been working on a patch that reduces the amount of
time printk spends with IRQs disabled.

David?

Cheers,

Jeff

Re: netconsole and a serial one

[Jon Smirl]

On Mon, Sep 22, 2008 at 9:19 AM, Jeff Moyer <jmoyer@redhat.com> wrote:
> "Jon Smirl" <jonsmirl@gmail.com> writes:
>
>> I have both netconsole and a serial console setup on an embedded
>> system. This is what I need for the boot process.
>>
>> After I'm booted I want to turn off printk() output to the serial
>> console but keep it going to netconsole.
>> I've been googling for an hour, how do I do this?
>>
>> I need to turn off the serial output after boot because serial
>> printk() disables interrupts for up to 1ms which is enough to cause me
>> to lose a lot of other interrupts.
>
> There's no way (that I know of) to do this presently.  However, I think
> David Howells has been working on a patch that reduces the amount of
> time printk spends with IRQs disabled.

I looked into interrupts and printk several years ago. If I recall
correctly printk() prints with interrupts off to ensure stability in
an OOPS situation.  Could serial printk() be changed to disable
interrupts only in the OOPS situation? It's all of the other
non-critical printks that are impacting my system.

-- 
Jon Smirl
jonsmirl@gmail.com