Hello,
On Thu, 14 Oct 2004 00:24:47 -0600 you wrote:
> I'm running ~x86 and 2.6.9-rc2-mm3 and something broke my system
> bootup.
>
> I got the issue resolved and so don't really care anymore, and since
> I'm running but I thought I'd let the community know for info's sake.
> (I know that I'm grateful for the fixes people post to this list.)
>
> After upgrading udev (and maybe other stuff), I get a critical error
> on startup after a bazillion "cannot change owner, object doesn't
> exist" or somethign like that for a whole bunch of /dev entries.
Your solution sounds a little bit complicated.
i also had some research on this. It turned out that in fact it's tar,
which is blocking. It seems that the new tar (1.14.90) tries to check if
permissions are set up right. It does that by opening the file for read
and again for write. This doesn't work on a lot of the dev files because
the kernel checks the device for existence when you open them.
Tar is failing with the mentioned error, which seems to be a little
misleading.
I resolved the issue by replacing tar by cpio (and bunzip2) in my
/sbin/rc.
I don't know right now where to file the bug, may be baselayout that
should use cpio instead of tar (it's also faster on my setups), may be
tar, which is doing the wrong thing when opening device nodes...
Hans-Werner
--
gentoo-user@gentoo.org mailing list
On Thu, 14 Oct 2004 00:24:47 -0600 you wrote:
> I'm running ~x86 and 2.6.9-rc2-mm3 and something broke my system
> bootup.
>
> I got the issue resolved and so don't really care anymore, and since
> I'm running but I thought I'd let the community know for info's sake.
> (I know that I'm grateful for the fixes people post to this list.)
>
> After upgrading udev (and maybe other stuff), I get a critical error
> on startup after a bazillion "cannot change owner, object doesn't
> exist" or somethign like that for a whole bunch of /dev entries.
Your solution sounds a little bit complicated.
i also had some research on this. It turned out that in fact it's tar,
which is blocking. It seems that the new tar (1.14.90) tries to check if
permissions are set up right. It does that by opening the file for read
and again for write. This doesn't work on a lot of the dev files because
the kernel checks the device for existence when you open them.
Tar is failing with the mentioned error, which seems to be a little
misleading.
I resolved the issue by replacing tar by cpio (and bunzip2) in my
/sbin/rc.
I don't know right now where to file the bug, may be baselayout that
should use cpio instead of tar (it's also faster on my setups), may be
tar, which is doing the wrong thing when opening device nodes...
Hans-Werner
--
gentoo-user@gentoo.org mailing list