Re: [PATCH] Store the relevant miscdevice in file->private_data inmisc_open()

[Greg KH]

On Thu, Nov 13, 2008 at 08:01:47PM -0800, Randy Dunlap wrote:
> Greg KH wrote:
> > On Fri, Nov 14, 2008 at 01:50:52PM +1100, Michael Ellerman wrote:
> >> On Thu, 2008-11-13 at 16:18 -0800, Greg KH wrote:
> >>> On Fri, Nov 14, 2008 at 10:54:41AM +1100, Michael Ellerman wrote:
> >>>> On Thu, 2008-11-13 at 09:31 -0800, Greg KH wrote: 
> >>>>> On Thu, Nov 13, 2008 at 03:49:50PM +1100, Michael Ellerman wrote:
> >>>>>> Currently it's not easy to share file_operations between multiple
> >>>>>> instances of a miscdevice. In order to do this, the device code needs to
> >>>>>> store a list of all it's miscdevice instances, and when fops->open() is
> >>>>>> called, search the list and find the right device based on the minor
> >>>>>> number.
> >>>>>>
> >>>>>> However the generic miscdevice code already has a list of miscdevices,
> >>>>>> and uses this to find the right device in misc_open(). If misc_open()
> >>>>>> would store the miscdevice it found in file->private_data, then the
> >>>>>> device code wouldn't need to worry about storing it's own separate list
> >>>>>> and searching that as well.
> >>>>>>
> >>>>>> The rest of the miscdevice code does not use file->private_data, so the
> >>>>>> device code is still free to use file->private_data for something else
> >>>>>> if it wants to.
> >>>>>>
> >>>>>> Signed-off-by: Michael Ellerman <michael@xxxxxxxxxxxxxx>
> >>>>> Do you have a follow-on patch for some misc device using code that would
> >>>>> take advantage of this change?
> >>>> Ah, good point. I do, but not for upstream :/
> >>> Hm, then I have to ask why should we take this change?
> >> Because it's seems like a good idea.
> > 
> > You know we don't make changes to core code for drivers that aren't in
> > the main tree, this is not a new thing...
> > 
> >>> And why would the driver not be availble for upstream to take?
> >> Because it's a hacky pile of crud, and it's for unreleased and
> >> non-existent hardware.
> > 
> > That's what the drivers/staging/ tree is for, send it on over to me and
> > I'll add it to that location.
> 
> for non-existent hardware??  that's not a good plan.

I read that as "not public yet" hardware.  If that is incorrect, Randy
is right, this isn't a good idea at all, sorry.

thanks,

greg k-h
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/