Re: Further madness in fs/partitions/check.c?

[James Bottomley <James.Bottomley@steeleye.com>]

> struct device contains a void * driver_data. It should certainly point
> to  a couple of bytes where the driver data was saved.

> In line 288, we have this:

> current_driverfs_dev->driver_data = (void *)__mkdev(hd->major,
> minor+part);

> What kind of pointer should we get here? ;-)

> Can the author please explain what was intented here?

This is transient code to save the device in the driver_data.  It is later 
picked back out at line 229.  It conforms to the old programmer principle that 
if you can always guarantee your data takes up less space than a pointer (on 
all archs), then you might as well just cast the data into the pointer instead 
of wasting a malloc for it.

The driverfs code is still in flux.  However, partition handling (if it 
belongs anywhere at all) will probably be in the unwritten class handlers and 
greatly tidied up.

The idea behind this code is to get a quick and dirty view of what partitions 
might be seen as in driverfs and thus to stimulate debate about how they 
should be done.

James