On Wed, Jun 13, 2012 at 08:24:12AM +0200, Paolo Bonzini wrote: > Il 13/06/2012 04:16, Dave Chinner ha scritto: > >> > + BUG_ON((mode & FALLOC_FL_PUNCH_HOLE) && (mode & FALLOC_FL_ZERO_RANGE)); > > Never put BUG_ON() or BUG() in XFS code that can return an error. > > Return EINVAL if we chose not to support it, and if it's really > > something we consider bad, emit a warning to syslog (i.e. > > xfs_warn()) and potentially add a ASSERT() case so that debug > > kernels will trip over it. Nobody should be panicing a production > > system just because a user supplied a set of incorrect syscall > > paramters.... > > I know, the BUG_ON() is because it is ruled out in VFS code. Of course > if I remove that code, this will not be a BUG_ON() anymore. If we put a BUG_ON() for every condition the VFS checked in every filesystem, we'd have so many BUG_ON checks we wouldn't be able to find the code. If it's banned at the VFS, there's no need to assert that inthe filesystem code.... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
