[Oracleasm-users] Compiling oracleasm from source ....

[Peter Larsen]

On Thu, 2006-06-15 at 17:33 -0700, Joel Becker wrote:
> On Thu, Jun 15, 2006 at 01:23:19AM -0400, Peter Larsen wrote:
> > I'm trying to get OracleAsm compiled under 2.6.16.9; as I need firewire 
> > etc. working I took a clean kernel source and did a configuration based 
> > on my RHAS4 + needed modules etc.
> 
> 	Um, I thought RHEL4 worked with firewire just fine.

Nope - ieee1394 is disabled/not installed as a module. I've had issues
before using the src RPMs for the kernel from redhat; I always build it
from scratch, using the .config from redhat.

> > However, oracleasm will not compile on this platform. First off, the 
> > ./configure fully ignores current kernel and just picks the original 
> > redhat supplied kernel. I fixed that by renaming /etc/redhat-release 
> > temporarily, but it seems that a structure is out of wack with the 
> > current kernel release? I get:
> 
> 	The configure will always default to the latest distro
> kernel-devel if you have it installed.  The
> --with-kernel=/usr/src/linux-2... option will force any kernel source
> path you like.

Ahh - great tip. Thank you.

> >    oracleasm-2.0.0/kernel/transaction_file.c: In function 'TA_write'
> >    oracleasm-2.0.0/kernel/transaction_file.c:55: error: structure has no 
> > memeber 'i_sem'
> 
> 	Yes, i_sem got renamed.  There are many parts of the kernel that
> have changed.  When SLES10/RHEL5 start coming out, you'll see oracleasm
> fix that up.  As oracleasm is an enterprise-system only thing, I don't
> have the time to keep it mainline happy.  Sorry :-(

gulp - I guess you're the developer here? <grumpy> well, I need to find
myself a kernel version that will do OCFS2 and oracleasm - what would
you suggest?

> > I wasn't able to get ocfs2 1.2.1 compiled under 2.6.15.1 where I had 
> > oracleasm working before. So I'm in a pinch here ;) What version of the 
> > kernel can I use - or is there a quick fix around this issue?
> 
> 	Again, I could swear RHEL4 firewire was supposed to work.
> Otherwise, you're probably going to have problems.  You could try
> patching oracleasm up to date, I guess.  the struct backing_dev bits
> have changed, i_sem->i_mutex and the associated mutex_lock() calls, etc.

Uhmmm - let me see if I have nothing else to do than venture into kernel
coding land this weekend :)

Regards
  Peter Larsen