Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 2 Jul 2021 16:59:54 +0200
From: John Wood <john.wood@....com>
To: Alexander Lobakin <alobakin@...me>
Cc: John Wood <john.wood@....com>, Kees Cook <keescook@...omium.org>,
	Jann Horn <jannh@...gle.com>, Jonathan Corbet <corbet@....net>,
	James Morris <jmorris@...ei.org>,
	"Serge E. Hallyn" <serge@...lyn.com>, Shuah Khan <shuah@...nel.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
	x86@...nel.org, "H. Peter Anvin" <hpa@...or.com>,
	Arnd Bergmann <arnd@...db.de>, Andi Kleen <ak@...ux.intel.com>,
	valdis.kletnieks@...edu,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Randy Dunlap <rdunlap@...radead.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-security-module@...r.kernel.org,
	linux-kselftest@...r.kernel.org, linux-arch@...r.kernel.org,
	linux-hardening@...r.kernel.org,
	kernel-hardening@...ts.openwall.com
Subject: Re: [PATCH v8 3/8] security/brute: Detect a brute force attack

Hi,

On Thu, Jul 01, 2021 at 11:55:14PM +0000, Alexander Lobakin wrote:
> Hi,
>
> From: John Wood <john.wood@....com>
> Date: Sat, 5 Jun 2021 17:04:00 +0200
>
> > +static int brute_task_execve(struct linux_binprm *bprm, struct file *file)
> > +{
> > +	struct dentry *dentry = file_dentry(bprm->file);
> > +	struct inode *inode = file_inode(bprm->file);
> > +	struct brute_stats stats;
> > +	int rc;
> > +
> > +	inode_lock(inode);
> > +	rc = brute_get_xattr_stats(dentry, inode, &stats);
> > +	if (WARN_ON_ONCE(rc && rc != -ENODATA))
> > +		goto unlock;
>
> I think I caught a problem here. Have you tested this with
> initramfs?

No, it has not been tested with initramfs :(

> According to init/do_mount.c's
> init_rootfs()/rootfs_init_fs_context(), when `root=` cmdline
> parameter is not empty, kernel creates rootfs of type ramfs
> (tmpfs otherwise).
> The thing about ramfs is that it doesn't support xattrs.

It is a known issue that systems without xattr support are not
suitable for Brute (there are a note in the documentation).
However, the purpose is not to panic the system :(

> I'm running this v8 on a regular PC with initramfs and having
> `root=` in cmdline, and Brute doesn't allow the kernel to run
> any init processes (/init, /sbin/init, ...) with err == -95
> (-EOPNOTSUPP) -- I'm getting a
>
> WARNING: CPU: 0 PID: 173 at brute_task_execve+0x15d/0x200
> <snip>
> Failed to execute /init (error -95)
>
> and so on (and a panic at the end).
>
> If I omit `root=` from cmdline, then the kernel runs init process
> just fine -- I guess because initramfs is then placed inside tmpfs
> with xattr support.
>
> As for me, this ramfs/tmpfs selection based on `root=` presence
> is ridiculous and I don't see or know any reasons behind that.
> But that's another story, and ramfs might be not the only one
> system without xattr support.
> I think Brute should have a fallback here, e.g. it could simply
> ignore files from xattr-incapable filesystems instead of such
> WARNING splats and stuff.

Ok, it seems reasonable to me: if the file system doesn't support
xattr, but Brute is enabled, Brute will do nothing and the system
will work normally.

I will work on it for the next version.
Thanks for the feedback.

John Wood

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.