On 5/14/2019 6:58 PM, Greg KH wrote:

On Tue, May 14, 2019 at 06:33:29PM +0200, Roberto Sassu wrote:

quoted

On 5/14/2019 5:19 PM, Andy Lutomirski wrote:

quoted

On Mon, May 13, 2019 at 5:47 AM Roberto Sassu [off-list ref] wrote:

quoted

On 5/13/2019 11:07 AM, Rob Landley wrote:

quoted

On 5/13/19 2:49 AM, Roberto Sassu wrote:

quoted

On 5/12/2019 9:43 PM, Arvind Sankar wrote:

quoted

On Sun, May 12, 2019 at 05:05:48PM +0000, Rob Landley wrote:

quoted

On 5/12/19 7:52 AM, Mimi Zohar wrote:

quoted

On Sun, 2019-05-12 at 11:17 +0200, Dominik Brodowski wrote:

quoted

On Thu, May 09, 2019 at 01:24:17PM +0200, Roberto Sassu wrote:

quoted

This proposal consists in marshaling pathnames and xattrs in a file called
.xattr-list. They are unmarshaled by the CPIO parser after all files have
been extracted.

Couldn't this parsing of the .xattr-list file and the setting of the xattrs
be done equivalently by the initramfs' /init? Why is kernel involvement
actually required here?

It's too late.  The /init itself should be signed and verified.

If the initramfs cpio.gz image was signed and verified by the extractor, how is
the init in it _not_ verified?

Ro

Wouldn't the below work even before enforcing signatures on external
initramfs:
1. Create an embedded initramfs with an /init that does the xattr
parsing/setting. This will be verified as part of the kernel image
signature, so no new code required.
2. Add a config option/boot parameter to panic the kernel if an external
initramfs attempts to overwrite anything in the embedded initramfs. This
prevents overwriting the embedded /init even if the external initramfs
is unverified.

Unfortunately, it wouldn't work. IMA is already initialized and it would
verify /init in the embedded initial ram disk.

So you made broken infrastructure that's causing you problems. Sounds unfortunate.

The idea is to be able to verify anything that is accessed, as soon as
rootfs is available, without distinction between embedded or external
initial ram disk.

Also, requiring an embedded initramfs for xattrs would be an issue for
systems that use it for other purposes.

quoted

quoted

The only reason why
opening .xattr-list works is that IMA is not yet initialized
(late_initcall vs rootfs_initcall).

Launching init before enabling ima is bad because... you didn't think of it?

No, because /init can potentially compromise the integrity of the
system.

I think Rob is right here.  If /init was statically built into the
kernel image, it has no more ability to compromise the kernel than
anything else in the kernel.  What's the problem here?

Right, the measurement/signature verification of the kernel image is
sufficient.

Now, assuming that we defer the IMA initialization until /init in the
embedded initramfs has been executed, the problem is how to handle
processes launched with the user mode helper or files directly read by
the kernel (if it can happen before /init is executed). If IMA is not
yet enabled, these operations will be performed without measurement and
signature verification.

If you really care about this, don't launch any user mode helper
programs (hint, you have the kernel option to control this and funnel
everything into one, or no, binaries).  And don't allow the kernel to
read any files either, again, you have control over this.

Or start IMA earlier if you need/want/care about this.

Yes, this is how it works now. It couldn't start earlier than
late_initcall, as it has to wait until the TPM driver is initialized.

Anyway, it is enabled at the time /init is executed. And this would be
an issue because launching /init and reading xattrs from /.xattr-list
would be denied (the signature is missing).

And /.xattr-list won't have a signature, if initramfs is generated
locally.

Roberto

-- 
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063
Managing Director: Bo PENG, Jian LI, Yanli SHI