Re: [Jfs-discussion] Do metapage pages need to be locked?

[Dave K. <sh...@au...>]

On Wednesday 24 April 2002 02:37 pm, Andrew Morton wrote:

> I guess all this stuff is still a couple of weeks away from
> readiness-for-inclusion.
>
> I'll do a first send of the core writeback changes tonight.
> JFS works OK with that code (well, it did in 2.5.8), but
> I suspect that it's subtly broken.  I'm no longer adding
> dirty buffers to inode.i_dirty_data_buffers, so the
> fsync_dirty_data_buffers() call won't actually write anything.
> I haven't tested recovery - probably this is broken by the
> change?

It does open up some opportunities for recovery to fail.

> Unlocking the metapage pages will fix all this up - the
> filemap_fdatasync() will do its work.

I can submit your patch to do this (cleaned up per your comments), or 
you can include it with your patches.  The problem I was fixing by 
locking the pages was not very pervasive.

> So if it's OK with you, I'll go ahead with the core writeback
> patch (possibly breaking recovery) and we work on the metapage
> writeback changes separately.

Go ahead with whatever patches you want to.  I don't think you'll break 
JFS that much by unlocking the metadata pages.  Leaving the recovery 
vulnerable at this stage would be acceptable as well.  We'll get it 
figured out soon enough.

By the way, I was able to run dbench with 50 clients on 2.5.9 on one 
processor without problems, on both UP and SMP kernels.  I guess I'll 
need to test on a real SMP box to reproduce the hang.
-- 
David Kleikamp
IBM Linux Technology Center