Hi folks, A couple of fixes things that Darrick and I noticed when using the iomap fiemap implementation. Darrick noticed that the attribute fork couldn't be mapped when updating his XFS scrubbing utility, and I noticed that the FIEMAP_FLAG_SYNC was being ignored by the code while looking at the attribute mapping issue. Both are regressions agains the old XFS fiemap implementation. The attribute mapping patch is the one that needs the most attention as it adds an iomap control flag and a new iomap internal error. The issue is that we don't know if there are attributes to map until we call down into the filesystem specific code, and at that point we need to tell iomap_apply() that there is nothing to map and so break out. This is done by returning -ENOENT, and then capturing that in iomap_fiemap() and treating it like we've reached the end of file. This causes fiemap to report the attribute map as having no extents. We also need to protect against write mappings being made against the attribute fork in XFS as it's not valid to write into the attribute extent map via this method. Hence we return -EINVAL if we detect an attempt to do so through the iomap control flags. I'm not sure it's the best way to solve the problem, but it does work. Better ideas are welcome... -Dave. _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
