Re: [patch] mlock-as-nonroot revisted

[Chris Wright <chrisw@osdl.org>]

* Andrea Arcangeli (andrea@suse.de) wrote:
> diff -purN linux-2.6.7/fs/hugetlbfs/inode.c linux/fs/hugetlbfs/inode.c
> --- linux-2.6.7/fs/hugetlbfs/inode.c    2004-07-29 11:36:55.744448953
> +0200
> +++ linux/fs/hugetlbfs/inode.c  2004-07-29 11:38:04.292595263 +0200
> @@ -722,7 +722,7 @@ struct file *hugetlb_zero_setup(size_t s
>         struct qstr quick_string;
>         char buf[16];
> 
> -       if (!capable(CAP_IPC_LOCK))
> +       if (!can_do_mlock())
>                 return ERR_PTR(-EPERM);
> 
>         if (!is_hugepage_mem_enough(size))
> 
> this breaks local security if you set the rlimit to 1 byte (well, 1 byte
> == disable_cap_mlock).

Right, that's true only for SHM_HUGETLB though, which uses
hugetlb_zero_setup.  And _that_ bit is what Rik's follow on patch is
trying to fix.  The normal hugetlbfs method using mmap() is unaffected,
AFAICT.  And has always been pretty open to using all of max_huge_pages.

thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net