Re: 2.5.70-mm7

[Christophe Saout <christophe@saout.de>]

Am Mit, 2003-06-11 um 01.12 schrieb Shane Shrybman:

> Yeah, I got the same. The message is something like 
> "ioctl cmd 2 No such address or device".

Ok, I think I found the problem.

In dm-ioctl.c in the function create this got changed:

-       int minor;
+       unsigned int minor = 0;

...

-       minor = (param->flags & DM_PERSISTENT_DEV_FLAG) ?
-               minor(to_kdev_t(param->dev)) : -1;
+       if (param->flags & DM_PERSISTENT_DEV_FLAG)
+               minor = minor(to_kdev_t(param->dev));

So, the variable minor is 0 now instead of -1 when the device shouldn't
be persistent. That's bad because 0 is a valid minor.

And now, in dm.c in alloc_dev, called by dm_create

        /* get a minor number for the dev */
-       minor = (minor < 0) ? next_free_minor() : specific_minor(minor);
-       if (minor < 0) {
+       r = (minor < 0) ? next_free_minor(&minor) :
specific_minor(minor);
+       if (r < 0) {

Here the minor is tested for being < 0 (which by the way can't be with
minor being unsigned, so this cleanup is bogus). The kernel always tries
to bind the new logical volume devices to the same minor, which of
course fails after the first one.

> Joe, do we need to upgrade some tools or something here?

No, because there are no new tools and the ioctl interface hasn't
changed (same dm-ioctl.h)

-- 
Christophe Saout <christophe@saout.de>