[bug#29232] [PATCH] gnu: qemu: Fix CVE-2017-{15038,15268,15289}.

[bug#29232] [PATCH] gnu: qemu: Fix CVE-2017-{15038,15268,15289}.

[Leo Famulari]

What do you think of fetching the patches like this, instead of copying
them into the Guix source tree?

* gnu/packages/virtualization.scm (qemu-patch): Use HTTPS.
(qemu)[source]: Use qemu-patch.
---
 gnu/packages/virtualization.scm | 31 +++++++++++++++++++++++--------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/gnu/packages/virtualization.scm b/gnu/packages/virtualization.scm
index 14b1dfbe0..2a2f41626 100644
--- a/gnu/packages/virtualization.scm
+++ b/gnu/packages/virtualization.scm
@@ -69,7 +69,7 @@
   (origin
     (method url-fetch)
     (uri (string-append
-          "http://git.qemu.org/?p=qemu.git;a=commitdiff_plain;h="
+          "https://git.qemu.org/?p=qemu.git;a=commitdiff_plain;h="
           commit))
     (sha256 sha256)
     (file-name file-name)))
@@ -78,13 +78,28 @@
   (package
     (name "qemu")
     (version "2.10.1")
-    (source (origin
-             (method url-fetch)
-             (uri (string-append "https://download.qemu.org/qemu-"
-                                 version ".tar.xz"))
-             (sha256
-              (base32
-               "1ahwl7r18iw2ds0q3c51nlivqsan9hcgnc8bbf9pv366iy81mm8x"))))
+    (source
+      (origin
+        (method url-fetch)
+        (uri (string-append "https://download.qemu.org/qemu-"
+                            version ".tar.xz"))
+        (patches
+          (list
+            (qemu-patch "7bd92756303f2158a68d5166264dc30139b813b6"
+                        "qemu-CVE-2017-15038.patch"
+                        (base32
+                         "0wpgf8ivjdbaihf2l7720h1fydh7kdl36wj2nchjd9irfkhw399q"))
+            (qemu-patch "a7b20a8efa28e5f22c26c06cd06c2f12bc863493"
+                        "qemu-CVE-2017-15268.patch"
+                        (base32
+                         "1adhwj91pmgbmdvyrkvslbfsyz7l00xdrr6vzps6s58q5idvdp79"))
+            (qemu-patch "eb38e1bc3740725ca29a535351de94107ec58d51"
+                        "qemu-CVE-2017-15289.patch"
+                        (base32
+                         "1zshrlzbwgwrsnimbq8kqr7injd65ncsr8a4lrmgyfv185ma4z8d"))))
+        (sha256
+         (base32
+          "1ahwl7r18iw2ds0q3c51nlivqsan9hcgnc8bbf9pv366iy81mm8x"))))
     (build-system gnu-build-system)
     (arguments
      '(;; Running tests in parallel can occasionally lead to failures, like:
-- 
2.15.0

[bug#29232] [PATCH] gnu: qemu: Fix CVE-2017-{15038,15268,15289}.

[Ludovic Courtès]

Hello,

Leo Famulari <leo@famulari.name> skribis:

> What do you think of fetching the patches like this, instead of copying
> them into the Guix source tree?

I think it’s OK.  If the Gitweb instance disappears, or if it changes
somehow, hopefully the patch itself will still have the same hash, so we
can always change to different URL or a local file.

> * gnu/packages/virtualization.scm (qemu-patch): Use HTTPS.
> (qemu)[source]: Use qemu-patch.

[…]

> +            (qemu-patch "7bd92756303f2158a68d5166264dc30139b813b6"
> +                        "qemu-CVE-2017-15038.patch"
> +                        (base32
> +                         "0wpgf8ivjdbaihf2l7720h1fydh7kdl36wj2nchjd9irfkhw399q"))
> +            (qemu-patch "a7b20a8efa28e5f22c26c06cd06c2f12bc863493"
> +                        "qemu-CVE-2017-15268.patch"
> +                        (base32
> +                         "1adhwj91pmgbmdvyrkvslbfsyz7l00xdrr6vzps6s58q5idvdp79"))
> +            (qemu-patch "eb38e1bc3740725ca29a535351de94107ec58d51"
> +                        "qemu-CVE-2017-15289.patch"
> +                        (base32
> +                         "1zshrlzbwgwrsnimbq8kqr7injd65ncsr8a4lrmgyfv185ma4z8d"))))

I trust these commits correspond to these CVEs.

Thanks,
Ludo’.

bug#29232: [PATCH] gnu: qemu: Fix CVE-2017-{15038,15268,15289}.

[Leo Famulari]

[-- Attachment #1: Type: text/plain, Size: 2180 bytes --]

On Thu, Nov 09, 2017 at 11:51:48PM +0100, Ludovic Courtès wrote:
> Hello,
> 
> Leo Famulari <leo@famulari.name> skribis:
> 
> > What do you think of fetching the patches like this, instead of copying
> > them into the Guix source tree?
> 
> I think it’s OK.  If the Gitweb instance disappears, or if it changes
> somehow, hopefully the patch itself will still have the same hash, so we
> can always change to different URL or a local file.
> 
> > * gnu/packages/virtualization.scm (qemu-patch): Use HTTPS.
> > (qemu)[source]: Use qemu-patch.
> 
> […]
> 
> > +            (qemu-patch "7bd92756303f2158a68d5166264dc30139b813b6"
> > +                        "qemu-CVE-2017-15038.patch"
> > +                        (base32
> > +                         "0wpgf8ivjdbaihf2l7720h1fydh7kdl36wj2nchjd9irfkhw399q"))
> > +            (qemu-patch "a7b20a8efa28e5f22c26c06cd06c2f12bc863493"
> > +                        "qemu-CVE-2017-15268.patch"
> > +                        (base32
> > +                         "1adhwj91pmgbmdvyrkvslbfsyz7l00xdrr6vzps6s58q5idvdp79"))
> > +            (qemu-patch "eb38e1bc3740725ca29a535351de94107ec58d51"
> > +                        "qemu-CVE-2017-15289.patch"
> > +                        (base32
> > +                         "1zshrlzbwgwrsnimbq8kqr7injd65ncsr8a4lrmgyfv185ma4z8d"))))
> 
> I trust these commits correspond to these CVEs.

Okay, I pushed adf7e69cab6180ef75360a1c0731c93f4bff2b18, which uses good
ol' annotated patch files instead.

Fetching the patches like this is too opaque. There's no *easy* way to
view the patches or figure out where they came from. The upstream
commits don't mention the CVE ID, and every interested person has to
re-do the work of corrolating the patch with the ID'd bug.

In practice, I think this extra works means that nobody will ever review
the patches or check that they correspond to a particular bug. Making
that easy is worth the extra bytes in our source tree.

Also I'm not confident that it will be easy to find bit-reproducible
patches in the future, whereas I think it will be easy to find the QEMU
tarballs and the patches from our Git repo.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[bug#29232] [PATCH] gnu: qemu: Fix CVE-2017-{15038,15268,15289}.

[Ludovic Courtès]

Hello,

Leo Famulari <leo@famulari.name> skribis:

> Okay, I pushed adf7e69cab6180ef75360a1c0731c93f4bff2b18, which uses good
> ol' annotated patch files instead.
>
> Fetching the patches like this is too opaque. There's no *easy* way to
> view the patches or figure out where they came from. The upstream
> commits don't mention the CVE ID, and every interested person has to
> re-do the work of corrolating the patch with the ID'd bug.
>
> In practice, I think this extra works means that nobody will ever review
> the patches or check that they correspond to a particular bug. Making
> that easy is worth the extra bytes in our source tree.

True.  I’m more inclined to skim over CVE patches that are inlined than
in this case.

> Also I'm not confident that it will be easy to find bit-reproducible
> patches in the future, whereas I think it will be easy to find the QEMU
> tarballs and the patches from our Git repo.

I’m a little bit more confident given that the Gitweb-generated patches
are merely raw commits, but I get your point.

Thanks!

Ludo’.