Device tree node to major/minor?

Grant Likely grant.likely at secretlab.ca
Thu Nov 22 02:47:58 EST 2012


On Tue, 20 Nov 2012 15:48:24 -0800, Simon Glass <sjg at chromium.org> wrote:
> Hi Grant,
> 
> On Tue, Nov 20, 2012 at 2:32 PM, Grant Likely <grant.likely at secretlab.ca> wrote:
> > On Tue, Nov 20, 2012 at 10:23 PM, Simon Glass <sjg at chromium.org> wrote:
> >> Hi,
> >>
> >> I hope this is a stupid question with an easy answer, but I cannot find it.
> >>
> >> I have a device tree node for an mmc block device and I want to use
> >> that block device from another driver. I have a phandle which lets me
> >> get the node of the mmc device, but I am not sure how to convert that
> >> into a block_device. In order to do so, I think I need a major/minor
> >> number. Of course the phandle might in fact point to a SCSI driver and
> >> I want that to work correctly also.
> >>
> >> I imagine I might be able to search through the wonders of sysfs in
> >> user space, but is there a better way?
> >
> > Do you /want/ to do it from userspace? What is your use case? Mounting
> > the rootfs?
> 
> The use case is storing some raw data on a block device from within a
> driver in the kernel. It is used to keep track of the verified boot
> state.
> 
> >
> > Regardless, userspace can monitor the uevents when devices are added
> > (that's what udev does) and watch for the full path of the node you
> > want in the uevent attribute. Then you can look for the child device
> > with the block major/minor numbers in it.
> 
> So is there a way to do this entirely in the kernel ex post? It might
> need to happen during kernel boot, before user space.

Yes, it is certainly doable within the kernel. First, you'll need to use
a notifier to get called back whenever a new device is created. Then
you'll need to look at the dev->of_node(->full_name) to see if it is the
node you actually want. You might need/want to resolve it from an alias
or something, but I presume you already have a way to find the
device_node before seaching for a struct device.

g.


More information about the devicetree-discuss mailing list