[Pkg-uml-devel] 1) noexec for shm 2) Is that a uml bug? 3) Is that rootstrap bug?

[Mattia Dongili]

On Sun, Nov 12, 2006 at 01:58:34AM +0200, shaulka at 012.net.il wrote:
> 1) noexec for shm:
>     =============
>   On Wednesday, November 8, 2006, Mattia Dongili 
> Subject: Re: [Pkg-uml-devel] Are you aware to 386945 and lkml 0609.2/1537.html
> 
> > The initscripts have been reverted to the old behaviour for
> > etch already but I think the package has yet to be migrated from sid.
> > 
> 
>    Still, this is only temporary:

Yes, but at least we have some time to discuss the thing upstream and
see if it's possible to fix the kernel there instead of having to
maintain a separate patch.

>   I only skimmed those bug reports. If I understood it correctly, the 
> noexec flag is claimed to be a security risk.

Yes, not a very strong argument, I remeber the discussion starting with
something like "there's no reason to have it mounted exec", not a real
known security hole.

> 2) Is the following a uml bug?
>    ======================
>   By  taking sysvinit from sid I managed to get rootstrap running. However
> now I believe my first problem is the following:
> 
> $ grep -A21 Failed rootstrap.log
> Failed to open 'root_fs', errno = 2
>  ubdb: unknown partition table
> VFS: Mounted root (hostfs filesystem) readonly.
> builder running...
> idr_remove called for id=2 which is not allocated.
> 0891ba10:  [<0805d048>] dump_stack+0x1c/0x20
[...]
> 
>   It looks as if the lines starting with `idr_remove' get repeated 2 more times
> (haven't carefully compared the text).

this should be harmless. Most probably due to the fact we are using a
python script as the init process.

> 3) Is the following a rootstrap bug? 
>    ==========================
>   I used rootstrap -o option. I do have more problems with uml,  such as setting
> the network. Still, I got a shell. However the shell prompt, `sh-3.1# ', as well as
> my keystrokes and the output of the commands were only printed to the log file.
> I was expecting it to appear on the tty too. Is that related to the previous oops?
> Is that a rootstrap bug?

yup. Yes, this is a rootstrap bug :)
We need to take more care with debug=1 and the -o option set.

-- 
mattia
:wq!