On Wed, 2019-10-16 at 01:40 -0700, Christoph Hellwig wrote: > On Wed, Oct 16, 2019 at 08:41:38AM +0800, Ian Kent wrote: > > +static int > > +xfs_validate_params( > > + struct xfs_mount *mp, > > + int dsunit, > > + int dswidth, > > + uint8_t iosizelog, > > + bool nooptions) > > Please add a refactor patch before this one to always set the stripe > unit / width / log I/O size values in the mount structure at parsing > time as suggested before. The will also remove the need for the > xfs_fs_context structure in the last patch. With the mount api it looks like this means adding the validation checks to xfs_parse_param(). That could be done for iosizelog without problem other than the check being redone for each instance of the "allocsize" option which is probably ok. But the dsunit and dswidth checks depend on one another so that check needs to be done after all the options have been handled. After the change to use the mount api the options string separation is done in the VFS and xfs_parse_param() called separately for each option. This is done so that the new system call fsconfig() can be implemented which will call the .parse_param() method separately for each option too. So it isn't known if both "sunit" and "swidth" have been given until aft er all the options have been looked at. Ian
