[bug#50698] [PATCH] WIP patches for recently-known hurd security vulnerabilities

[Ludovic Courtès]

Hi Maxime,

Maxime Devos <maximedevos@telenet.be> skribis:

> I've tried to patch the glibc package for the problems noted at
> <https://lists.gnu.org/archive/html/bug-hurd/2021-08/msg00007.html>;.
>
> I've found two recent patches (glibc-hurd-proc-reauth.patch and
> glibc-hurd-sendmsg-SCM_CREDS.patch) that appeared relevant.  I tried
> to patch our glibc package with those patches.
>
> The modified tarball builds fine for --system=x86_64-linux, but not
> for --system=i586-gnu (tested with ./pre-inst-env guix build hello
> --system=i586-gnu).  Any idea what's happening here?

Thanks for looking into it!

> From cdf38fbfcba4c87777d7ba2175f08e877dafe86a Mon Sep 17 00:00:00 2001
> From: Maxime Devos <maximedevos@telenet.be>
> Date: Mon, 13 Sep 2021 11:23:21 +0200
> Subject: [PATCH] WIP gnu: glibc: New security patches.
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> The existence of the vulnerabilities was noted at
> <https://lists.gnu.org/archive/html/bug-hurd/2021-08/msg00007.html>.
>
> TODO: check if these are all necessary packages for glibc.
> TODO: why does the glibc tarball build for --system=x86_64-linux but not
> for --system=i586-gnu?
>
> Build error:
> ‘patching file hurd/hurdinit.c
> Hunk #1 FAILED at 177.
> 1 out of 1 hunk FAILED -- saving rejects to file hurd/hurdinit.c.rej’
>
> but this file isn't modified by the new patches!
>
> * gnu/local.mk (dist_patch_DATA): Register new patches.
> * gnu/packages/base.scm (glibc)[replacement]: Register replacement.
>   (glibc/fixed): New variable.
> * gnu/packages/patches/glibc-hurd-proc-reauth.patch: New file.
> * gnu/packages/patches/glibc-hurd-sendmsg-SCM_CREDS.patch.

[...]

> --- a/gnu/packages/base.scm
> +++ b/gnu/packages/base.scm
> @@ -706,6 +706,7 @@ the store.")
>    (package
>     (name "glibc")
>     (version "2.31")
> +   (replacement glibc/fixed)
>     (source (origin
>              (method url-fetch)
>              (uri (string-append "mirror://gnu/glibc/glibc-" version 
> ".tar.xz"))
> @@ -966,6 +967,12 @@ with the Linux kernel.")
>     (license lgpl2.0+)
>     (home-page "https://www.gnu.org/software/libc/";)))
>  
> +(define glibc/fixed
> +  (package-with-extra-patches
> +   glibc
> +   (search-patches "glibc-hurd-sendmsg-SCM_CREDS.patch"
> +                   "glibc-hurd-proc-reauth.patch")))

Instead of a replacement, which makes no sense on GNU/Linux, could you
add a conditional phase for (hurd-target?) that applies the patches?

(On ‘core-updates’ (or ‘-frozen’?) we will apply patches
unconditionally.)

Not answering your initial question, but maybe the problem will vanish
if you do things this way, who knows.  :-)

> +++ b/gnu/packages/patches/glibc-hurd-proc-reauth.patch
> @@ -0,0 +1,114 @@
> +Index: glibc-2.31/hurd/hurdsig.c

Please add a comment explaining what this patch does, what its status
is, with a link to upstream discussions.

Thank you!

Ludo’.