removal of EXPORT_SYMBOL(insert_resource)?

removal of EXPORT_SYMBOL(insert_resource)?

[Kumar Gala]

I have a situation that I believe warrants leaving insert_resource as an 
exported API.

I've got a bus implementation that it done as a module.  While I'm more 
than happy to provide this bus implementation to be included in the 
mainline, I dont think it makes much sense to do so.  The code is only 
useful to an extremely small handful of people.  If we want to clutter the 
kernel with it I'm happy to provide a patch for it.

The situation I have is a FPGA connect over PCI.  The FPGA implements a 
number of different "functions" but uses PCI more like an SoC bus than a 
true PCI device.  Anyways, in some discussions with gregkh, it was 
suggested the best thing was to create a new bus type that the "fpga" 
drivers would bind to.

I use insert_resource to handle registering the MMIO regions for each 
device (similar to how platform devices are registered).

- kumar

Re: removal of EXPORT_SYMBOL(insert_resource)?

[Greg KH]

On Thu, Mar 02, 2006 at 01:31:08PM -0600, Kumar Gala wrote:
> I have a situation that I believe warrants leaving insert_resource as an 
> exported API.
> 
> I've got a bus implementation that it done as a module.  While I'm more 
> than happy to provide this bus implementation to be included in the 
> mainline, I dont think it makes much sense to do so.  The code is only 
> useful to an extremely small handful of people.  If we want to clutter the 
> kernel with it I'm happy to provide a patch for it.

Please do, keeping code outside the kernel makes it _very_ hard on you.
It makes it easier if everything is in-the-tree, as you know.

Hell, we have two whole x86 subarchs with only 4 machines each in
existance, a simple bus is nothing :)

> The situation I have is a FPGA connect over PCI.  The FPGA implements a 
> number of different "functions" but uses PCI more like an SoC bus than a 
> true PCI device.  Anyways, in some discussions with gregkh, it was 
> suggested the best thing was to create a new bus type that the "fpga" 
> drivers would bind to.
> 
> I use insert_resource to handle registering the MMIO regions for each 
> device (similar to how platform devices are registered).

All the better reason to get it into the tree...

thanks,

greg k-h

Re: removal of EXPORT_SYMBOL(insert_resource)?

[Adrian Bunk]

Kumar,

any news regarding a submission of your code?

cu
Adrian


On Fri, Mar 03, 2006 at 04:59:35PM -0800, Greg KH wrote:
> On Thu, Mar 02, 2006 at 01:31:08PM -0600, Kumar Gala wrote:
> > I have a situation that I believe warrants leaving insert_resource as an 
> > exported API.
> > 
> > I've got a bus implementation that it done as a module.  While I'm more 
> > than happy to provide this bus implementation to be included in the 
> > mainline, I dont think it makes much sense to do so.  The code is only 
> > useful to an extremely small handful of people.  If we want to clutter the 
> > kernel with it I'm happy to provide a patch for it.
> 
> Please do, keeping code outside the kernel makes it _very_ hard on you.
> It makes it easier if everything is in-the-tree, as you know.
> 
> Hell, we have two whole x86 subarchs with only 4 machines each in
> existance, a simple bus is nothing :)
> 
> > The situation I have is a FPGA connect over PCI.  The FPGA implements a 
> > number of different "functions" but uses PCI more like an SoC bus than a 
> > true PCI device.  Anyways, in some discussions with gregkh, it was 
> > suggested the best thing was to create a new bus type that the "fpga" 
> > drivers would bind to.
> > 
> > I use insert_resource to handle registering the MMIO regions for each 
> > device (similar to how platform devices are registered).
> 
> All the better reason to get it into the tree...
> 
> thanks,
> 
> greg k-h