From: Carlos O'Donell <carlos@redhat.com>
To: Florian Weimer <fweimer@redhat.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>, libc-stable@sourceware.org
Subject: Re: Backporting CVE-2016-10739
Date: Tue, 01 Jan 2019 00:00:00 -0000 [thread overview]
Message-ID: <2e644830-b506-f5b6-e020-99fc9ee9b94f@redhat.com> (raw)
In-Reply-To: <877eefmk3z.fsf@oldenburg2.str.redhat.com>
On 2/4/19 2:25 PM, Florian Weimer wrote:
> * Carlos O'Donell:
>
>> On 2/4/19 11:18 AM, Florian Weimer wrote:
>>> * Carlos O'Donell:
>>>
>>>> On 2/4/19 10:44 AM, Florian Weimer wrote:
>>>>> * Carlos O'Donell:
>>>>>
>>>>>>> +#include <arpa/inet.h>
>>>>>>> +
>>>>>>
>>>>>> Please add a comment explaining why this is here.
>>>>>
>>>>> You mean like this?
>>>>>
>>>>> /* Obtain the prototype for __inet_aton_exact. */
>>>>
>>>> It should reference the bug or CVE to document the intent
>>>> of the changes.
>>>>
>>>> Post v3 and I'll sign off?
>>>
>>> This approach does not actually work because copying a prototype this
>>> way and adding a hidden visibility attribute does not actually make the
>>> symbol hidden. The patch below however has the desired effect, mainly
>>> because interposition no longer happens and the __inet_aton_exact_hidden
>>> function is not added to the dynamic symbol table of the nscd
>>> executable.
>>>
>>> I suspect I would have had to use __attribute__ ((visibility
>>> ("hidden"))) directly because we define attribute_hidden thusly:
>>>
>>> #if defined SHARED || defined LIBC_NONSHARED \
>>> || (BUILD_PIE_DEFAULT && IS_IN (libc))
>>> # define attribute_hidden __attribute__ ((visibility ("hidden")))
>>> #else
>>> # define attribute_hidden
>>> #endif
>>>
>>> I can post yet another patch which uses real hidden visibility and
>>> avoids the symbol redirect.
>>
>> This version is a little hack-ish, but I don't object to it. It does
>> the job.
>>
>> It is certainly a minimal set of changes, and that makes it quite
>> useful for the release branch.
>>
>> I don't think you need to go any further unless you think the version
>> using real hidden visibility is all that much better?
>
> Well, at least it does not have the completely ineffective
> attribute_hidden. 8-)
I'm just providing high-level review here, I haven't tested these patches,
I rely on you for that :-}
> Patch below. What do you think?
This looks good to me, you make direct use of "__attribute__ ((visibility ("hidden")))"
in an exceptional case, and that's fine.
If this becomes less rare for some reason we might want a libc-symbols.h macro to define
something that expresses the intent of the hidden visibility e.g. attr_dup_sym_hidden.
It's rare we have a fix that needs a new private interface, and that we can also
just dup the code to fix the problem. Most likely I expect we *can't* dup the code
and so need the new private interface to fix the bug, or we rework the fix entirely.
> This one gets the symbol visibility correct.
OK.
> $ eu-readelf -s ../build/nscd/nscd-inet_addr.o | grep inet_aton_exact
> 22: 0000000000000150 56 FUNC GLOBAL HIDDEN 1 __inet_aton_exact
> $ eu-readelf -s ../build/nscd/gai.o | grep inet_aton_exact
> 99: 0000000000000000 0 NOTYPE GLOBAL HIDDEN UNDEF __inet_aton_exact
That's good.
>
> And there's no __inet_aton_exact symbol in nscd. I think this is fairly
> standard usage of a hidden symbol, so renaming the symbol should not be
> necessary.
Perfect, and interposition is also not possible.
> Thanks,
> Florian
>
> nscd: Do not use __inet_aton_exact@GLIBC_PRIVATE [BZ #20018]
>
> This commit avoids referencing the __inet_aton_exact@GLIBC_PRIVATE
> symbol from nscd. In master, the separately-compiled getaddrinfo
> implementation in nscd needs it, however such an internal ABI change
> is not desirable on a release branch if it can be avoided.
>
> 2019-02-04 Florian Weimer <fweimer@redhat.com>
>
> [BZ #20018]
> nscd: Do not rely on new GLIBC_PRIVATE ABI after CVE-2016-10739 fix.
> * nscd/nscd-inet_addr.c: New file. Build resolv/inet_addr.c for
> nscd, without public symbols.
> * nscd/Makefile (nscd-modules): Add it.
> * nscd/gai.c: Include <arpa/inet.h> and change visibility of
> __inet_aton_exact.
OK for release/2.28/master.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
> diff --git a/nscd/Makefile b/nscd/Makefile
> index b713a84c49..eb23c01a39 100644
> --- a/nscd/Makefile
> +++ b/nscd/Makefile
> @@ -36,7 +36,7 @@ nscd-modules := nscd connections pwdcache getpwnam_r getpwuid_r grpcache \
> getsrvbynm_r getsrvbypt_r servicescache \
> dbg_log nscd_conf nscd_stat cache mem nscd_setup_thread \
> xmalloc xstrdup aicache initgrcache gai res_hconf \
> - netgroupcache
> + netgroupcache nscd-inet_addr
OK.
>
> ifeq ($(build-nscd)$(have-thread-library),yesyes)
>
> diff --git a/nscd/gai.c b/nscd/gai.c
> index f57f396f57..68a4abd30e 100644
> --- a/nscd/gai.c
> +++ b/nscd/gai.c
> @@ -33,6 +33,12 @@
> #define __getifaddrs getifaddrs
> #define __freeifaddrs freeifaddrs
>
> +/* We do not want to export __inet_aton_exact. Get the prototype and
> + change its visibility to hidden. */
> +#include <arpa/inet.h>
> +__typeof__ (__inet_aton_exact) __inet_aton_exact
> + __attribute__ ((visibility ("hidden")));
OK.
> +
> /* We are nscd, so we don't want to be talking to ourselves. */
> #undef USE_NSCD
>
> diff --git a/nscd/nscd-inet_addr.c b/nscd/nscd-inet_addr.c
> new file mode 100644
> index 0000000000..f366b9567d
> --- /dev/null
> +++ b/nscd/nscd-inet_addr.c
> @@ -0,0 +1,32 @@
> +/* Legacy IPv4 text-to-address functions. Version for nscd.
> + Copyright (C) 2019 Free Software Foundation, Inc.
OK.
> + This file is part of the GNU C Library.
> +
> + The GNU C Library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + The GNU C Library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with the GNU C Library; if not, see
> + <http://www.gnu.org/licenses/>. */
> +
> +#include <arpa/inet.h>
> +
> +/* We do not want to export __inet_aton_exact. Get the prototype and
> + change the visibility to hidden. */
> +#include <arpa/inet.h>
> +__typeof__ (__inet_aton_exact) __inet_aton_exact
> + __attribute__ ((visibility ("hidden")));
OK.
> +
> +/* Do not provide definitions of the public symbols exported from
> + libc. */
> +#undef weak_alias
> +#define weak_alias(from, to)
OK.
> +
> +#include <resolv/inet_addr.c>
>
--
Cheers,
Carlos.
next prev parent reply other threads:[~2019-02-04 19:58 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-01-01 0:00 Aurelien Jarno
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell [this message]
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Carlos O'Donell
2019-01-01 0:00 ` Aurelien Jarno
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Aurelien Jarno
2019-01-01 0:00 ` Florian Weimer
2019-01-01 0:00 ` Aurelien Jarno
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=2e644830-b506-f5b6-e020-99fc9ee9b94f@redhat.com \
--to=carlos@redhat.com \
--cc=aurelien@aurel32.net \
--cc=fweimer@redhat.com \
--cc=libc-stable@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).