[bug#31487] [PATCH] gnu: Add upx.

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

* gnu/packages/compression.scm (ucl): New variable.
* gnu/packages/compression.scm (upx): New variable.
---
 gnu/packages/compression.scm | 73 ++++++++++++++++++++++++++++++++++++
 1 file changed, 73 insertions(+)

diff --git a/gnu/packages/compression.scm b/gnu/packages/compression.scm
index 562a2bf8b..b0d7cd971 100644
--- a/gnu/packages/compression.scm
+++ b/gnu/packages/compression.scm
@@ -2151,3 +2151,76 @@ with @code{deflate} but offers more dense compression.
 
 The specification of the Brotli Compressed Data Format is defined in RFC 7932.")
     (license license:expat)))
+
+(define-public ucl
+  (package
+    (name "ucl")
+    (version "1.03")
+    (source (origin
+             (method url-fetch)
+             (uri (string-append "http://www.oberhumer.com/opensource/"
+                                 name "/download/" name "-" version ".tar.gz"))
+             (sha256
+              (base32
+               "0j036lkwsxvm15gr29n8wn07cqq79dswjs9k54939ms5zngjjrdq"))))
+    (build-system gnu-build-system)
+    (home-page "http://www.oberhumer.com/opensource/ucl/")
+    (synopsis "Portable lossless data compression library")
+    (description "UCL implements a number of compression algorithms that
+achieve an excellent compression ratio while allowing *very* fast
+decompression.  Decompression requires no additional memory.
+
+UCL is an OpenSource re-implementation of some NRV compression algorithms.
+
+As compared to LZO, the UCL algorithms achieve a better compression ratio but
+decompression is a little bit slower.  See below for some rough timings.")
+    (license license:gpl2)))
+
+(define-public upx
+  (package
+    (name "upx")
+    (version "3.94")
+    (source (origin
+             (method url-fetch)
+             (uri (string-append "https://github.com/upx/upx/releases/download/v"
+                                 version "/" name "-" version "-src.tar.xz"))
+             (sha256
+              (base32
+               "08anybdliqsbsl6x835iwzljahnm9i7v26icdjkcv33xmk6p5vw1"))))
+    (build-system gnu-build-system)
+    (native-inputs `(("zlib" ,zlib)
+                     ("perl" ,perl)
+                     ("ucl" ,ucl)))
+    (arguments
+     `(#:make-flags
+       (list "all"
+             ;; CHECK_WHITESPACE does not seem to work.
+             ;; See https://git.archlinux.org/svntogit/community.git/tree/trunk/PKGBUILD?h=packages/upx.
+             "CHECK_WHITESPACE=true")
+       #:phases
+       (modify-phases %standard-phases
+         (delete 'configure)
+         (delete 'check)
+         (delete 'install)
+         (add-before 'build 'patch-exec-bin-sh
+           (lambda _
+             (substitute* (find-files "Makefile")
+               (("/bin/sh") (which "sh")))
+             (substitute* "src/Makefile"
+               (("/bin/sh") (which "sh")))
+             #t))
+         (add-after 'build 'install-upx
+           (lambda* (#:key outputs #:allow-other-keys)
+             (let* ((out (assoc-ref outputs "out"))
+                   (bin (string-append out "/bin")))
+               (mkdir-p bin)
+               (copy-file "src/upx.out" (string-append bin "/upx")))
+             #t))
+         )))
+    (home-page "https://upx.github.io/")
+    (synopsis "The Ultimate Packer for eXecutables")
+    (description "UPX is an advanced executable file compressor.  UPX will
+typically reduce the file size of programs and DLLs by around 50%-70%, thus
+reducing disk space, network load times, download times and other distribution
+and storage costs.")
+    (license license:gpl2)))
-- 
2.17.0

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

[-- Attachment #1: Type: text/plain, Size: 73 bytes --]


Forgot to add my name to the copyright list.

-- 
Pierre Neidhardt

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[bug#31487] [PATCH] gnu: Add upx.

[Ludovic Courtès]

Hello,

Pierre Neidhardt <ambrevar@gmail.com> skribis:

> * gnu/packages/compression.scm (ucl): New variable.
> * gnu/packages/compression.scm (upx): New variable.

I committed both as separate patches (the convention is to have one
patch per package), slightly changed descriptions to remove “marketing
speak”, and changed licenses to ‘gpl2+’ after checking the source file
headers.

There’s one issue left though:

  $ ./pre-inst-env guix lint upx
  gnu/packages/compression.scm:2179:2: upx@3.94: probably vulnerable to CVE-2017-15056, CVE-2017-16869

Could you check whether patches are available for these?  Better be safe
than sorry!

Thank you,
Ludo’.

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

[-- Attachment #1: Type: text/plain, Size: 677 bytes --]


Ludovic Courtès <ludo@gnu.org> writes:

> There’s one issue left though:
>
>   $ ./pre-inst-env guix lint upx
>   gnu/packages/compression.scm:2179:2: upx@3.94: probably vulnerable to CVE-2017-15056, CVE-2017-16869
>
> Could you check whether patches are available for these?  Better be safe
> than sorry!

Indeed they are.
They are not on the master branch though, only devel I think.
So what's the protocol here?  Shall we cherry-pick the fixing commits or
get latest devel?

-- 
Pierre Neidhardt

The day advanced as if to light some work of mine; it was morning,
and lo! now it is evening, and nothing memorable is accomplished.
		-- H.D. Thoreau

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[bug#31487] [PATCH] gnu: Add upx.

[Ludovic Courtès]

Hi Pierre,

Pierre Neidhardt <ambrevar@gmail.com> skribis:

> Ludovic Courtès <ludo@gnu.org> writes:
>
>> There’s one issue left though:
>>
>>   $ ./pre-inst-env guix lint upx
>>   gnu/packages/compression.scm:2179:2: upx@3.94: probably vulnerable to CVE-2017-15056, CVE-2017-16869
>>
>> Could you check whether patches are available for these?  Better be safe
>> than sorry!
>
> Indeed they are.
> They are not on the master branch though, only devel I think.
> So what's the protocol here?  Shall we cherry-pick the fixing commits or
> get latest devel?

Yes.  You can add them as individual patches (see commit
aa8ac0294421d465f60e18c8271f971ec8407a95 for an example); as usual, make
sure each patch starts with a few lines explaining what the patch does
and where it comes from (you can take the commit log for that plus a
repo URL, for instance.)

Then you can check that ‘guix lint upx’ is happy.

TIA!

Ludo’.

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

[-- Attachment #1: Type: text/plain, Size: 207 bytes --]


The relevant issues:

- https://github.com/upx/upx/issues/146
- https://github.com/upx/upx/pull/190

Both CVEs were rejected, so I guess there is no need to include a patch.

-- 
Pierre Neidhardt

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[bug#31487] [PATCH] gnu: Add upx.

[Ludovic Courtès]

Pierre Neidhardt <ambrevar@gmail.com> skribis:

> The relevant issues:
>
> - https://github.com/upx/upx/issues/146
> - https://github.com/upx/upx/pull/190

Hmm I see that:

  https://github.com/upx/upx/issues/128
  corresponds to:
  https://nvd.nist.gov/vuln/detail?vulnId=CVE%2D%32%30%31%37%2D%31%35%30%35%36

and:

  https://nvd.nist.gov/vuln/detail?vulnId=CVE%2D%32%30%31%37%2D%31%36%38%36%39
  corresponds to:
  https://github.com/upx/upx/issues/146

The latter (CVE-2017-16869) is marked as “disputed” above, and I would
agree with the arguments of the UPX maintainers.

The authors did not react to the former (CVE-2017-15056, crash when
reading ELF files), other than by fixing it, but it does look similar in
spirit.

What about adding a patch for CVE-2017-15056 since it would at least fix
a concrete bug?

CVE-2017-16869 is also a bug but it concerns Mach-O files, which are
much less of a concern for our users I suppose.  Patching it wouldn’t
hurt either, but you could also add a ‘lint-hidden-cve’ property for
CVE-2017-16869 with a comment.

TIA,
Ludo’.

[bug#31487] [PATCH] gnu: Add upx.

[Ludovic Courtès]

Ping!  :-)

ludo@gnu.org (Ludovic Courtès) skribis:

> Pierre Neidhardt <ambrevar@gmail.com> skribis:
>
>> The relevant issues:
>>
>> - https://github.com/upx/upx/issues/146
>> - https://github.com/upx/upx/pull/190
>
> Hmm I see that:
>
>   https://github.com/upx/upx/issues/128
>   corresponds to:
>   https://nvd.nist.gov/vuln/detail?vulnId=CVE%2D%32%30%31%37%2D%31%35%30%35%36
>
> and:
>
>   https://nvd.nist.gov/vuln/detail?vulnId=CVE%2D%32%30%31%37%2D%31%36%38%36%39
>   corresponds to:
>   https://github.com/upx/upx/issues/146
>
> The latter (CVE-2017-16869) is marked as “disputed” above, and I would
> agree with the arguments of the UPX maintainers.
>
> The authors did not react to the former (CVE-2017-15056, crash when
> reading ELF files), other than by fixing it, but it does look similar in
> spirit.
>
> What about adding a patch for CVE-2017-15056 since it would at least fix
> a concrete bug?
>
> CVE-2017-16869 is also a bug but it concerns Mach-O files, which are
> much less of a concern for our users I suppose.  Patching it wouldn’t
> hurt either, but you could also add a ‘lint-hidden-cve’ property for
> CVE-2017-16869 with a comment.
>
> TIA,
> Ludo’.

[bug#31487] [PATCH] gnu: upx: Fix CVE-2017-15056.

[Pierre Neidhardt]

* gnu/packages/patches/upx-protect-against-bad-crafted-input.patch: New file.
* gnu/packages/compression.scm (upx)[source]: Use it.
---
 gnu/packages/compression.scm                  |  8 +-
 ...px-protect-against-bad-crafted-input.patch | 96 +++++++++++++++++++
 2 files changed, 103 insertions(+), 1 deletion(-)
 create mode 100644 gnu/packages/patches/upx-protect-against-bad-crafted-input.patch

diff --git a/gnu/packages/compression.scm b/gnu/packages/compression.scm
index 8f062049a..0be7962b3 100644
--- a/gnu/packages/compression.scm
+++ b/gnu/packages/compression.scm
@@ -2209,7 +2209,8 @@ decompression is a little bit slower.")
                                  version "/" name "-" version "-src.tar.xz"))
              (sha256
               (base32
-               "08anybdliqsbsl6x835iwzljahnm9i7v26icdjkcv33xmk6p5vw1"))))
+               "08anybdliqsbsl6x835iwzljahnm9i7v26icdjkcv33xmk6p5vw1"))
+             (patches (search-patches "upx-protect-against-bad-crafted-input.patch"))))
     (build-system gnu-build-system)
     (native-inputs `(("perl" ,perl)
                      ("ucl" ,ucl)))
@@ -2241,6 +2242,11 @@ decompression is a little bit slower.")
              #t))
          )))
     (home-page "https://upx.github.io/")
+    ;; CVE-16869 is about Mach-O files which is not of a big concern for Guix.
+    ;; See https://github.com/upx/upx/issues/146 and
+    ;; https://nvd.nist.gov/vuln/detail?vulnId=CVE%2D%32%30%31%37%2D%31%36%38%36%39.
+    ;; The issue will be fixed after version 3.94.
+    (properties `((lint-hidden-cve . ("CVE-2017-16869"))))
     (synopsis "Compression tool for executables")
     (description
      "The Ultimate Packer for eXecutables (UPX) is an executable file
diff --git a/gnu/packages/patches/upx-protect-against-bad-crafted-input.patch b/gnu/packages/patches/upx-protect-against-bad-crafted-input.patch
new file mode 100644
index 000000000..525980e73
--- /dev/null
+++ b/gnu/packages/patches/upx-protect-against-bad-crafted-input.patch
@@ -0,0 +1,96 @@
+From 3e0c2966dffb5dadb512a476ef4be3d0cc51c2be Mon Sep 17 00:00:00 2001
+From: Pierre Neidhardt <ambrevar@gmail.com>
+Date: Sat, 16 Jun 2018 16:35:00 +0200
+Subject: [PATCH] Protect against bad crafted input
+
+Also check for wrap-around when checking oversize involving e_shoff and e_shnum.
+
+raised by https://github.com/upx/upx/pull/190
+	modified:   p_lx_elf.cpp
+---
+ src/p_lx_elf.cpp | 30 ++++++++++++++++++++++++++++++
+ 1 file changed, 30 insertions(+)
+
+diff --git a/src/p_lx_elf.cpp b/src/p_lx_elf.cpp
+index 822a7652..41e805ee 100644
+--- a/src/p_lx_elf.cpp
++++ b/src/p_lx_elf.cpp
+@@ -235,8 +235,17 @@ PackLinuxElf32::PackLinuxElf32help1(InputFile *f)
+         sz_phdrs = 0;
+         return;
+     }
++    if (0==e_phnum) throwCantUnpack("0==e_phnum");
+     e_phoff = get_te32(&ehdri.e_phoff);
++    unsigned const last_Phdr = e_phoff + e_phnum * sizeof(Elf32_Phdr);
++    if (last_Phdr < e_phoff || (unsigned long)file_size < last_Phdr) {
++        throwCantUnpack("bad e_phoff");
++    }
+     e_shoff = get_te32(&ehdri.e_shoff);
++    unsigned const last_Shdr = e_shoff + e_shnum * sizeof(Elf32_Shdr);
++    if (last_Shdr < e_shoff || (unsigned long)file_size < last_Shdr) {
++        throwCantUnpack("bad e_shoff");
++    }
+     sz_phdrs = e_phnum * e_phentsize;
+ 
+     if (f && Elf32_Ehdr::ET_DYN!=e_type) {
+@@ -599,8 +608,17 @@ PackLinuxElf64::PackLinuxElf64help1(InputFile *f)
+         sz_phdrs = 0;
+         return;
+     }
++    if (0==e_phnum) throwCantUnpack("0==e_phnum");
+     e_phoff = get_te64(&ehdri.e_phoff);
++    upx_uint64_t const last_Phdr = e_phoff + e_phnum * sizeof(Elf64_Phdr);
++    if (last_Phdr < e_phoff || (unsigned long)file_size < last_Phdr) {
++        throwCantUnpack("bad e_phoff");
++    }
+     e_shoff = get_te64(&ehdri.e_shoff);
++    upx_uint64_t const last_Shdr = e_shoff + e_shnum * sizeof(Elf64_Shdr);
++    if (last_Shdr < e_shoff || (unsigned long)file_size < last_Shdr) {
++        throwCantUnpack("bad e_shoff");
++    }
+     sz_phdrs = e_phnum * e_phentsize;
+ 
+     if (f && Elf64_Ehdr::ET_DYN!=e_type) {
+@@ -3763,6 +3781,9 @@ void PackLinuxElf64::pack4(OutputFile *fo, Filter &ft)
+ 
+ void PackLinuxElf64::unpack(OutputFile *fo)
+ {
++    if (e_phoff != sizeof(Elf64_Ehdr)) {// Phdrs not contiguous with Ehdr
++        throwCantUnpack("bad e_phoff");
++    }
+     unsigned const c_phnum = get_te16(&ehdri.e_phnum);
+     upx_uint64_t old_data_off = 0;
+     upx_uint64_t old_data_len = 0;
+@@ -3828,6 +3849,9 @@ void PackLinuxElf64::unpack(OutputFile *fo)
+     unsigned total_out = 0;
+     unsigned c_adler = upx_adler32(NULL, 0);
+     unsigned u_adler = upx_adler32(NULL, 0);
++    if ((MAX_ELF_HDR - sizeof(Elf64_Ehdr))/sizeof(Elf64_Phdr) < u_phnum) {
++        throwCantUnpack("bad compressed e_phnum");
++    }
+ 
+     // Packed ET_EXE has no PT_DYNAMIC.
+     // Packed ET_DYN has original PT_DYNAMIC for info needed by rtld.
+@@ -4383,6 +4407,9 @@ Elf64_Sym const *PackLinuxElf64::elf_lookup(char const *name) const
+ 
+ void PackLinuxElf32::unpack(OutputFile *fo)
+ {
++    if (e_phoff != sizeof(Elf32_Ehdr)) {// Phdrs not contiguous with Ehdr
++        throwCantUnpack("bad e_phoff");
++    }
+     unsigned const c_phnum = get_te16(&ehdri.e_phnum);
+     unsigned old_data_off = 0;
+     unsigned old_data_len = 0;
+@@ -4449,6 +4476,9 @@ void PackLinuxElf32::unpack(OutputFile *fo)
+     unsigned total_out = 0;
+     unsigned c_adler = upx_adler32(NULL, 0);
+     unsigned u_adler = upx_adler32(NULL, 0);
++    if ((MAX_ELF_HDR - sizeof(Elf32_Ehdr))/sizeof(Elf32_Phdr) < u_phnum) {
++        throwCantUnpack("bad compressed e_phnum");
++    }
+ 
+     // Packed ET_EXE has no PT_DYNAMIC.
+     // Packed ET_DYN has original PT_DYNAMIC for info needed by rtld.
+-- 
+2.17.0
+
-- 
2.17.0

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

[-- Attachment #1: Type: text/plain, Size: 220 bytes --]


Ludovic Courtès <ludo@gnu.org> writes:

> Ping!  :-)

Sorry for the long delay.  I've just sent a patch.

I'm having issues with my development setup so I haven't been able to
test it.

--
Pierre Neidhardt

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

[bug#31487] [PATCH] gnu: Add upx.

[Pierre Neidhardt]

[-- Attachment #1: Type: text/plain, Size: 155 bytes --]


I've finally resolved issues with my Guix setup and I've successfully
tested the patched version of upx.
Feel free to merge.

-- 
Pierre Neidhardt

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

bug#31487: [PATCH] gnu: upx: Fix CVE-2017-15056.

[Ludovic Courtès]

Hello,

Pierre Neidhardt <ambrevar@gmail.com> skribis:

> * gnu/packages/patches/upx-protect-against-bad-crafted-input.patch: New file.
> * gnu/packages/compression.scm (upx)[source]: Use it.

I renamed the patch so that it includes the CVE id, added it to
gnu/local.mk, and committed.

Thanks!

Ludo’.