linux-kernel - Re: [PATCH 4.19 105/116] bpf: Explicitly memset the bpf

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <20200402185320.GA8077@duo.ucw.cz>
Date:   Thu, 2 Apr 2020 20:53:21 +0200
From:   Pavel Machek <pavel@...x.de>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc:     linux-kernel@...r.kernel.org, stable@...r.kernel.org,
        Maciej Żenczykowski <maze@...gle.com>,
        John Stultz <john.stultz@...aro.org>,
        Alexander Potapenko <glider@...gle.com>,
        Alistair Delva <adelva@...gle.com>,
        Daniel Borkmann <daniel@...earbox.net>,
        Yonghong Song <yhs@...com>
Subject: Re: [PATCH 4.19 105/116] bpf: Explicitly memset the bpf_attr
 structure

Hi!

> From: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
> 
> commit 8096f229421f7b22433775e928d506f0342e5907 upstream.
> 
> For the bpf syscall, we are relying on the compiler to properly zero out
> the bpf_attr union that we copy userspace data into. Unfortunately that
> doesn't always work properly, padding and other oddities might not be
> correctly zeroed, and in some tests odd things have been found when the
> stack is pre-initialized to other values.
> 
> Fix this by explicitly memsetting the structure to 0 before using
> it.

Is not that a gcc bug? I mean, that's seriously unhelpful behaviour
from security perspective.

Is there any reason to believe this is not causing problems elsewhere?

$ grep -ri "= {}" . | wc -l
2152

I'm pretty sure many of these are before return to userspace... I
picked one at random:

./drivers/media/cec/cec-api.c-static long cec_adap_g_caps(struct cec_adapter *adap,
./drivers/media/cec/cec-api.c-                      struct cec_caps __user *parg)
./drivers/media/cec/cec-api.c-{
./drivers/media/cec/cec-api.c:  struct cec_caps caps = {};
./drivers/media/cec/cec-api.c-
./drivers/media/cec/cec-api.c-  strscpy(caps.driver, adap->devnode.dev.parent->driver->name,
./drivers/media/cec/cec-api.c-          sizeof(caps.driver));
./drivers/media/cec/cec-api.c-  strscpy(caps.name, adap->name, sizeof(caps.name));
./drivers/media/cec/cec-api.c-  caps.available_log_addrs = adap->available_log_addrs;
./drivers/media/cec/cec-api.c-  caps.capabilities = adap->capabilities;
./drivers/media/cec/cec-api.c-  caps.version = LINUX_VERSION_CODE;
./drivers/media/cec/cec-api.c-  if (copy_to_user(parg, &caps, sizeof(caps)))
./drivers/media/cec/cec-api.c-          return -EFAULT;
./drivers/media/cec/cec-api.c-  return 0;
./drivers/media/cec/cec-api.c-}

Should we fix gcc, instead?

Best regards,
								Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Download attachment "signature.asc" of type "application/pgp-signature" (196 bytes)