[6.8-rc1 Regression] Unable to exec apparmor_parser from virt-aa-helper

[Tetsuo Handa]

On 2024/01/27 20:00, Tetsuo Handa wrote:
> On 2024/01/27 16:04, Tetsuo Handa wrote:
>> If we can accept revival of security_bprm_free(), we can "get rid of current->in_execve flag"
>> and "stop saving things across two *independent* execve() calls".
> 
> Oops, I found a bug in TOMOYO (and possibly in AppArmor as well).
> TOMOYO has to continue depending on current->in_execve flag even if
> security_bprm_free() is revived.

No. We can "get rid of current->in_execve flag" and "stop saving things across
two *independent* execve() calls".

> @@ -327,9 +322,13 @@ static int tomoyo_file_fcntl(struct file *file, unsigned int cmd,
>   */
>  static int tomoyo_file_open(struct file *f)
>  {
> -       /* Don't check read permission here if called from execve(). */
> -       /* Illogically, FMODE_EXEC is in f_flags, not f_mode. */
> -       if (f->f_flags & __FMODE_EXEC)
> +       /*
> +        * Don't check read permission here if called from execve() for
> +        * the first time of that execve() request, for execute permission
> +        * will be checked at tomoyo_bprm_check_security() with argv/envp
> +        * taken into account.
> +        */
> +       if (current->in_execve && !tomoyo_task(current)->old_domain_info)

Since "f->f_flags & __FMODE_EXEC" == "current->in_execve", TOMOYO can continue using
"f->f_flags & __FMODE_EXEC", provided that tomoyo_task(current)->old_domain_info is
reset to NULL via security_bprm_free() callback when previous execve() request failed.

That is, if security_bprm_free() is revived, we can also get rid of current->in_execve.