Re: [PATCH 4/4] tune2fs: Add support for orphan_file feature

[Jan Kara <jack@suse.cz>]

On Fri 22-05-15 10:18:42, Darrick J. Wong wrote:
> On Fri, May 22, 2015 at 01:28:57PM +0200, Jan Kara wrote:
> > @@ -1970,6 +2011,31 @@ static int parse_extended_opts(ext2_filsys fs, const char *opts)
> >  				continue;
> >  			}
> >  			ext_mount_opts = strdup(arg);
> > +		} else if (!strcmp(token, "orphan_file_size")) {
> > +			__u64 size;
> > +
> > +			if (!arg) {
> > +				r_usage++;
> > +				continue;
> > +			}
> > +			size = strtoul(arg, &p, 0);
> 
> Would be nice if I could supply units, e.g. orphan_file_size = 128K here.

I don't mind doing this but do we have precedens in any option? Because
e.g. journal size is just a number (in megabytes), stripe options are in fs
blocks. For orphan file megabytes are too coarse. I can make the unit
fs-blocks, or kbytes, or leave it at bytes, or allow specifying units but
some consistency would be good. Any opinion?

> > +			if (*p) {
> > +				fprintf(stderr,
> > +					_("Invalid size of orphan file %s\n"),
> > +					arg);
> > +				r_usage++;
> > +				continue;
> > +			}
> > +			if (size < EXT4_MIN_ORPHAN_FILE_SIZE) {
> > +				fprintf(stderr,
> > +					_("Orphan file is too small. Minimum "
> > +					  "size is %u\n"),
> > +					EXT4_MIN_ORPHAN_FILE_SIZE);
> > +				r_usage++;
> > +				continue;
> > +			}
> > +			orphan_file_blocks = (size + fs->blocksize - 1) /
> > +						fs->blocksize;
> >  		} else
> >  			r_usage++;
> >  	}
> > @@ -2921,6 +2987,17 @@ retry_open:
> >  		if (rc)
> >  			goto closefs;
> >  	}
> > +	if (orphan_file_blocks) {
> 
> If someone specifies -E orphan_file_size=NNN -O ^orphan_file, does this have
> the effect of erasing and recreating the orphan file?

Yes, although that's a side-effect of the option parsing :) Intended use
is that you first disable the orphan_file feature and then enable it again
with new size.

									Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR