real-time device BUG

[Eric Sandeen]

On 1/29/16 5:40 PM, Ross Zwisler wrote:
> On Fri, Jan 29, 2016 at 04:05:31PM -0700, Ross Zwisler wrote:
>> Today I was trying to get real-time devices working with XFS, and both a
>> manual pwrite() to a real-time device and xfs_rtcp end up with a lockdep spat
>> followed by a kernel BUG.  This behavior is consistent between kernel versions
>> v4.4 and v4.5-rc1.  I initially found this using some pmem block devices, but
>> it also happens with BRD.
> 
> This also happens with an mmap on the real-time file followed by a write.
> Essentially I think that any path that tries to allocate a block for the real
> time file his this same BUG.

Looks like an oversight from

commit eef334e5776c8ef547ada4cec17549929fe590b4
Author: Christoph Hellwig <hch at infradead.org>
Date:   Fri Dec 6 12:30:17 2013 -0800

    xfs: assert that we hold the ilock for extent map access
    
    Make sure that xfs_bmapi_read has the ilock held in some way, and that
    xfs_bmapi_write, xfs_bmapi_delay, xfs_bunmapi and xfs_iread_extents are
    called with the ilock held exclusively.
    
    Signed-off-by: Christoph Hellwig <hch at lst.de>
    Reviewed-by: Dave Chinner <dchinner at redhat.com>
    Signed-off-by: Ben Myers <bpm at sgi.com>

I think all we have to do is lock it for read here, I'll send a patch and
Dave or Christoph can smack me down if I'm wrong ;)