src/patches/xzgrep-ZDI-CAN-16587.patch

   1 From 69d1b3fc29677af8ade8dc15dba83f0589cb63d6 Mon Sep 17 00:00:00 2001
   2 From: Lasse Collin <lasse.collin@tukaani.org>
   3 Date: Tue, 29 Mar 2022 19:19:12 +0300
   4 Subject: [PATCH] xzgrep: Fix escaping of malicious filenames (ZDI-CAN-16587).
   5
   6 Malicious filenames can make xzgrep to write to arbitrary files
   7 or (with a GNU sed extension) lead to arbitrary code execution.
   8
   9 xzgrep from XZ Utils versions up to and including 5.2.5 are
  10 affected. 5.3.1alpha and 5.3.2alpha are affected as well.
  11 This patch works for all of them.
  12
  13 This bug was inherited from gzip's zgrep. gzip 1.12 includes
  14 a fix for zgrep.
  15
  16 The issue with the old sed script is that with multiple newlines,
  17 the N-command will read the second line of input, then the
  18 s-commands will be skipped because it's not the end of the
  19 file yet, then a new sed cycle starts and the pattern space
  20 is printed and emptied. So only the last line or two get escaped.
  21
  22 One way to fix this would be to read all lines into the pattern
  23 space first. However, the included fix is even simpler: All lines
  24 except the last line get a backslash appended at the end. To ensure
  25 that shell command substitution doesn't eat a possible trailing
  26 newline, a colon is appended to the filename before escaping.
  27 The colon is later used to separate the filename from the grep
  28 output so it is fine to add it here instead of a few lines later.
  29
  30 The old code also wasn't POSIX compliant as it used \n in the
  31 replacement section of the s-command. Using \<newline> is the
  32 POSIX compatible method.
  33
  34 LC_ALL=C was added to the two critical sed commands. POSIX sed
  35 manual recommends it when using sed to manipulate pathnames
  36 because in other locales invalid multibyte sequences might
  37 cause issues with some sed implementations. In case of GNU sed,
  38 these particular sed scripts wouldn't have such problems but some
  39 other scripts could have, see:
  40
  41     info '(sed)Locale Considerations'
  42
  43 This vulnerability was discovered by:
  44 cleemy desu wayo working with Trend Micro Zero Day Initiative
  45
  46 Thanks to Jim Meyering and Paul Eggert discussing the different
  47 ways to fix this and for coordinating the patch release schedule
  48 with gzip.
  49 ---
  50  src/scripts/xzgrep.in | 20 ++++++++++++--------
  51  1 file changed, 12 insertions(+), 8 deletions(-)
  52
  53 diff --git a/src/scripts/xzgrep.in b/src/scripts/xzgrep.in
  54 index b180936..e5186ba 100644
  55 --- a/src/scripts/xzgrep.in
  56 +++ b/src/scripts/xzgrep.in
  57 @@ -180,22 +180,26 @@ for i; do
  58           { test $# -eq 1 || test $no_filename -eq 1; }; then
  59        eval "$grep"
  60      else
  61 +      # Append a colon so that the last character will never be a newline
  62 +      # which would otherwise get lost in shell command substitution.
  63 +      i="$i:"
  64 +
  65 +      # Escape & \ | and newlines only if such characters are present
  66 +      # (speed optimization).
  67        case $i in
  68        (*'
  69  '* | *'&'* | *'\'* | *'|'*)
  70 -        i=$(printf '%s\n' "$i" |
  71 -            sed '
  72 -              $!N
  73 -              $s/[&\|]/\\&/g
  74 -              $s/\n/\\n/g
  75 -            ');;
  76 +        i=$(printf '%s\n' "$i" | LC_ALL=C sed 's/[&\|]/\\&/g; $!s/$/\\/');;
  77        esac
  78 -      sed_script="s|^|$i:|"
  79 +
  80 +      # $i already ends with a colon so don't add it here.
  81 +      sed_script="s|^|$i|"
  82
  83        # Fail if grep or sed fails.
  84        r=$(
  85          exec 4>&1
  86 -        (eval "$grep" 4>&-; echo $? >&4) 3>&- | sed "$sed_script" >&3 4>&-
  87 +        (eval "$grep" 4>&-; echo $? >&4) 3>&- |
  88 +            LC_ALL=C sed "$sed_script" >&3 4>&-
  89        ) || r=2
  90        exit $r
  91      fi >&3 5>&-
  92 --
  93 2.35.1
  94