Re: RFC: Put printk buffer in video ram

[Maxim Levitsky <maximlevitsky@gmail.com>]

On Sun, 2009-11-22 at 14:15 +0200, Maxim Levitsky wrote: 
> On Sun, 2009-11-22 at 09:25 +0100, Ingo Molnar wrote: 
> > * Frederic Weisbecker <fweisbec@gmail.com> wrote:
> > 
> > > On Sun, Nov 22, 2009 at 04:05:06AM +0200, Maxim Levitsky wrote:
> > > > After doing some successful debugging by placing printk buffer in video
> > > > ram, here I publish cleaned version of it.
> > > > 
> > > > I discovered that on my system video ram isn't cleared on reboot, and I
> > > > took advantage of that by placing printk buffer directly there.
> > > > This allows to capture oopses/panicks almost from everywhere.
> > > > It is also very simple to setup.

Few more thoughts:

First of all, if I implement this as a console driver, I won't be able
to capture all kernel log, but only from the point I register the
console.
My implementation, also isn't set up very early, but I copy already
written log to the memory buffer.

Problem of automatic unregistration isn't a problem, its just a feature,
I could skip using.

Secondary, I found out that system ram isn't cleared ether on my
notebook, however I found out that I can't reserve small memory range
with 'memmap' kernel parameter. System just panics.
I used for reference 'memmap=2M$0x60000000'
My system has 2GB of memory.

I could reserve the memory same way using the code, but it would
probably fail in same way. Also I could try early reservations, but I am
not sure I can reserve _any_ address range. I test this.

I think that patches to put printk buffer at predefined area in system
ram were once posted on LKML.



And now, to be honest, lets talk about existing solutions:

1 - kdump/kexec - not a bad thing, but it has some disadvantages, namely
I need to compile seperate kernel, there is a probility of new kernel
not booting up (in fact I once installed kexec-tools, and it made system
reboot using kexec, and I remember a hang because of that...)
However, this approach doesn't need any 'help' from BIOS.

2 - mtdoops + phram - There is a console driver that logs into mtd
device. I don't know how it selects one. 
It is possible to create a 'fake' mtd device that will span a range in
physical memory.
I couldn't make this work, it has problems with early output, etc..


Best regards,
Maxim Levitsky