Re: Patch 1300 & rpm issue with 1.3.0

[Russell Cattelan <cattelan@xfs.org>]

On Mon, 2003-08-25 at 10:50, Axel Thimm wrote:
> On Sun, Aug 24, 2003 at 11:27:05AM -0500, Eric Sandeen wrote:
> > On Sun, 24 Aug 2003, Kai Leibrandt wrote:
> > > I just got the 2.4.20-19.9_XFS1.3.0.src.rpm from oss.sgi.com and after 
> > > building, installing and rebooting I noticed the same issue with rpm that 
> > > hit me with 2.4.20-18.9... Of course after adding patch 1300 back in all is 
> > > fine again, so just a quick question; is this going to be the standard way 
> > > the .src.rpm are packaged, or am I doing something else wrong?
> > 
> > Boy, that's an annoying bug...  it's somewhere in the guts of Red Hat's
> > kernel + nptl patches + O_DIRECT + rpm.  I think that Red Hat will eventually
> > have a new version of RPM that works with this kernel.  In the meantime,
> > I'd either:
> > 
> > a) rebuild with patch 1300 in place, if you don't care about using O_DIRECT
> > or
> > b) set up an alias for "rpm" to prefix it with LD_ASSUME_KERNEL=2.2.5
> 
> The good news is that the current rawhide kernel has commented out
> that patch (named 1140 there). Unfortunatley there is no changelog
> entry as to why and how this was done.
> 
> > > Also, is there a howto or readme somewhere where I can find out how to build 
> > > my own .src.rpm kernel packages from the redhat errata and the XFS patches?
> > 
> > No docs... best documentation is probably the spec file itself.  Which xfs
> > patches, and which Red Hat kernel packages, do you want to combine?
> 
> I played a bit with rawhide (which is still the same as severn's, I
> think). Other than the _nolock patch the rest looked trivial. I
> haven't got a complete build yet, but it looked quite promising (I'm
> on vacation, so it will take me some time). Whoever cares to look at
> the current bits can contact me in private.
Try using this change from the 2.4.22 tree 
http://xfs.org:8090/linux-2.4+xfs/diffs/fs/xfs/linux/xfs_lrw.c@1.167?nav=index.html|src/|src/fs|src/fs/xfs|src/fs/xfs/linux|hist/fs/xfs/linux/xfs_lrw.c

I think the RH beta tree has do_generic_file_write at this point so you
should be able to call it directly and drop the whole _nolock change all
together.
You might need to export do_generic_file_write ... can't remember off
hand.

> 
> I was told to better skip 1.3.0 and go straight for the CVS bits for
> 2.4.22, but I wouldn't like to desync from the XFS release points.