On Tue, 2011-02-22 at 13:08 -0500, Laszlo Ersek wrote:
> On 02/22/11 18:44, Daniel Stodden wrote:
>
> > I'm pretty sure minors> 256 date way before 2.6.32. Here's the module
> > init fragment from blktap2, replacing the register_chrdev() call:
> >
> > [...]
>
> That's about what I was thinking of in [0]. It also sets the cdev's owner
> manually. However, register_chrdev() also does this:
>
> 225 kobject_set_name(&cdev->kobj, "%s", name);
> 226 for (s = strchr(kobject_name(&cdev->kobj),'/'); s; s =
> strchr(s, '/'))
> 227 *s = '!';
>
> I reckon we can ignore the 's,/,!,g' replacement, since the name is fixed
> "blktap" (or "blktap2"). But the kobject name doesn't appear to be set in the
> first place.
> Is that no problem? If not, I'd just omit it from blktap as well.
I remember I browsed a whole lot of code back then when I wrote that.
All gone now :)
After looking again, I think the way this works was that the kobj map
holds the blktap2 class device, nowhere a cdev->kobj, after
device_register (blktap2/sysfs.c) actually placed a node for that minor.
Before then, only the range is allocated, but won't show up in sysfs. So
I guess cdev->kobj is a somewhat anonymous refcount bag only.
# l /sys/dev/char/* | grep blktap
lrwxrwxrwx 1 root root 0 Feb 22 10:24 /sys/dev/char/10:61 ->
../../class/misc/blktap-control
lrwxrwxrwx 1 root root 0 Feb 22 10:24 /sys/dev/char/254:0 ->
../../class/blktap2/blktap0
lrwxrwxrwx 1 root root 0 Feb 22 10:24 /sys/dev/char/254:1 ->
../../class/blktap2/blktap1
lrwxrwxrwx 1 root root 0 Feb 22 10:24 /sys/dev/char/254:2 ->
../../class/blktap2/blktap2
Compare this to the number of tty mappings you'll probably find on your
box in the same directory.
Daniel
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home