Re: [PATCH] loop.c: kernel_thread() retval check

[Solar Designer <solar@openwall.com>]

On Sun, Aug 20, 2006 at 09:21:48AM +0200, Willy Tarreau wrote:
> I still remembered this problem being discussed, and finally found
> the thread :
> 
>   http://lkml.org/lkml/2003/11/14/55

I was not aware that this had been discussed before.  Bernhard (in the
old LKML posting above) seems to imply that having kernel_thread()
itself not fail on ptrace would be a sufficient fix, which I don't agree
with.  There may be other reasons for kernel_thread() to fail, such as
the kernel running out of resources; with OpenVZ, kernel_thread() is not
allowed from within VEs.

> In fact, no code was proposed and 2.6 got fixed later, then stopped
> using kernel_thread() so nearly nobody might have noticed it :
> 
>   http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff_plain;h=3e88c17d404c5787afd5bd1763380317f5ccbf84;hp=22e6c1b39c648850438decd491f62d311800c7db

I'm afraid that this does not properly clean things up on error.  I just
had a look at linux-2.6.17.9/drivers/block/loop.c - it still uses
kernel_thread() and has the same "goto out_putf" on error return from
kernel_thread(), which appears to not clean things up.

Alexander