linux-kernel - Re: [PATCH] ima: Fix undefined arch_ima_get

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20211215160345.GF3786@linux-l9pv.suse>
Date:   Thu, 16 Dec 2021 00:03:45 +0800
From:   joeyli <jlee@...e.com>
To:     Takashi Iwai <tiwai@...e.de>, Mimi Zohar <zohar@...ux.ibm.com>
Cc:     Mimi Zohar <zohar@...ux.ibm.com>,
        Dmitry Kasatkin <dmitry.kasatkin@...il.com>,
        linux-integrity@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] ima: Fix undefined arch_ima_get_secureboot() and co

Hi Takashi, Mimi,

On Tue, Dec 14, 2021 at 04:58:47PM +0100, Takashi Iwai wrote:
> On Tue, 14 Dec 2021 16:31:21 +0100,
> Mimi Zohar wrote:
> > 
> > Hi Takashi,
> > 
> > On Mon, 2021-12-13 at 17:11 +0100, Takashi Iwai wrote:
> > > Currently arch_ima_get_secureboot() and arch_get_ima_policy() are
> > > defined only when CONFIG_IMA is set, and this makes the code calling
> > > those functions without CONFIG_IMA failing.  Although there is no such
> > > in-tree users, but the out-of-tree users already hit it.
> > > 
> > > Move the declaration and the dummy definition of those functions
> > > outside ifdef-CONFIG_IMA block for fixing the undefined symbols.
> > > 
> > > Signed-off-by: Takashi Iwai <tiwai@...e.de>
> > 
> > Before lockdown was upstreamed, we made sure that IMA and lockdown
> > could co-exist.  This patch makes the stub functions available even
> > when IMA is not configured.  Do the remaining downstream patches
> > require IMA to be disabled or can IMA co-exist?
> 
> I guess Joey (Cc'ed) can explain this better.  AFAIK, currently it's
> used in a part of MODSIGN stuff in SUSE kernels, and it's calling
> unconditionally this function for checking whether the system is with
> the Secure Boot or not.
>

Actually in downstream code, I used arch_ima_get_secureboot() in
load_uefi_certs() to prevent the mok be loaded when secure boot is
disabled. Because the security of MOK relies on secure boot.

The downstream code likes this:

/* the MOK and MOKx can not be trusted when secure boot is disabled */
-      if (!efi_enabled(EFI_SECURE_BOOT))
+      if (!arch_ima_get_secureboot())
		return 0;

The old EFI_SECURE_BOOT bit can only be available on x86_64, so I switch
the code to to arch_ima_get_secureboot() for cross-architectures and sync
with upstream api.

The load_uefi.c depends on CONFIG_INTEGRITY but not CONFIG_IMA. So
load_uefi.c still be built when CONFIG_INTEGRITY=y and CONFIG_IMA=n.
Then "implicit declaration of function 'arch_ima_get_secureboot'" is
happened.

Thanks a lot!
Joey Lee