Re: [PATCH] xfs: byte range buffer dirty region tracking

[Dave Chinner <david@fromorbit.com>]

On Thu, Feb 01, 2018 at 03:22:58PM -0800, Darrick J. Wong wrote:
> On Fri, Feb 02, 2018 at 10:16:47AM +1100, Dave Chinner wrote:
> > On Thu, Feb 01, 2018 at 12:35:26PM -0800, Darrick J. Wong wrote:
> > > FWIW I also ran straight into this when I applied it for giggles and ran
> > > xfstests -g quick (generic/001 blew up):
> > 
> > I must have screwed up the forward port worse than usual - the
> > conflicts with the xfs_buf_log_item typedef removal were pretty
> > extensive.
> 
> Ah, sorry about that.  I'd thought it was just the xfs_buf rename. :/

Not your fault at all, Darrick!

I only complained about the xfs_buf typedef because it would cause
merge problems for ~80% of the patches in my current dev tree. This
was the only patch that the xfs_buf_log_item typedef removal
affected - more were affected by the trivial b_fspriv to b_log_item
changeover - and I figured that pain was worth it to get rid of
another typedef....

> > > [   31.909228] ================================================================================
> > > [   31.911258] BUG: unable to handle kernel NULL pointer dereference at 00000000000000a0
> > > [   31.912375] IP: xfs_buf_item_init+0x33/0x350 [xfs]
> > 
> > Hmmmm - I'm seeing that on my subvol smoke test script but not
> > elsewhere. I've been looking through the subvol code to try to find
> > this, maybe it's not the subvol code.  What mkfs parameters where
> > you using?
> 
> mkfs.xfs -m rmapbt=1,reflink=1 -i sparse=1 /dev/pmem0

OK, nothing unusual, though I haven't been using sparse=1 recently.
I'll get onto it....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com