Re: [evlog-dev] evlog module

Steve Woodruff wrote:
> 
> Hi Larry,
> 
> Addressing your points:
> 
> 1) i think *I* could live without the early boot
>    messages.  our use of evlog is based on having
>    one of our evlog plugins receive events and
>    then do "stuff".  During boot time we can't
>    do that "stuff" since it requires our other
>    user-space applications to be up and running.
>    Having evlog as a module would give an option
>    for those like me who can afford to miss those
>    early messages.

It certainly would be possible to implement as a module,
but read my next answer.  Perhaps some of my assumptions
are incorrect.

> 
> 2) I'm not sure i understand.  Where are we defining
>    new logging functions?  In evlog?  User apps?

Based on your next statement below, I think you are 
assuming that an evlog kernel module could work without
_any_ changes in bzImage at all.  So then how could you
hook into the printk() message stream for forwarding 
printk without patching printk.c ? 

If you decided you want evl_writek() for logging binary 
data, don't you need to include it in a header file that's
picked-up when you create bzImage ?

OTOH, if your driver was built as a module, and called
a function in the evlog module (which was loaded first),
then this scenario just might work.  Is this what you have
in mind ?  

> 
> From what our hardware vendor tells us, RH simply
> will not support any kernel that they didn't
> compile.  

Yes, I understand.  We indirectly have ongoing discussions
with RH.  

Thanks,
Larry


-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com