s3

   1 #!/bin/bash
   2 # basic amazon s3 operations
   3 # Licensed under the terms of the GNU GPL v2
   4 # Copyright 2007 Victor Lowther <victor.lowther@gmail.com>
   5
   6 CURL=/home/mwolson_admin/bin/curl
   7 HMAC=$(dirname $0)/s3-hmac
   8
   9 # print a message and bail
  10 die() {
  11   echo $*
  12   exit 1
  13 }
  14
  15 # check to see if the variable name passed exists and holds a value.
  16 # Die if it does not.
  17 check_or_die() {
  18   [[ ${!1} ]] || die "Environment variable ${1} is not set."
  19 }
  20
  21 # check to see if we have all the needed S3 variables defined.
  22 # Bail if we do not.
  23 check_s3() {
  24   local sak x
  25   for x in S3_ACCESS_KEY_ID S3_SECRET_ACCESS_KEY; do
  26     check_or_die ${x};
  27   done
  28   [[ -f ${S3_SECRET_ACCESS_KEY} ]] || die "S3_SECRET_ACCESS_KEY must point to a file!"
  29   sak="$(wc -c "${S3_SECRET_ACCESS_KEY}")"
  30   (( ${sak%%[!0-9 ]*} == 40 )) || \
  31     die "S3 Secret Access Key is not exactly 40 bytes long.  Please fix it."
  32 }
  33 # check to see if our external dependencies exist
  34 check_dep() {
  35   local res=0
  36   while [[ $# -ne 0 ]]; do
  37     which "${1}" >& /dev/null || { res=1; echo "${1} not found."; }
  38     shift
  39   done
  40   (( res == 0 )) || die "aborting."
  41 }
  42
  43 check_hmac() {
  44   if test ! -f $HMAC || test ! -x $HMAC; then
  45     die "hmac script not found or not executable."
  46   fi
  47 }
  48
  49 check_deps() {
  50   check_dep openssl date cat grep
  51   check_hmac
  52   check_s3
  53 }
  54
  55 urlenc() {
  56   # $1 = string to url encode
  57   # output is on stdout
  58   # we don't urlencode everything, just enough stuff.
  59   echo -n "${1}" |
  60   sed 's/%/%25/g
  61        s/ /%20/g
  62        s/#/%23/g
  63        s/\$/%24/g
  64        s/\&/%26/g
  65        s/+/%2b/g
  66        s/,/%2c/g
  67        s/:/%3a/g
  68        s/;/%3b/g
  69        s/?/%3f/g
  70        s/@/%40/g
  71        s/       /%09/g'
  72 }
  73
  74 xmldec() {
  75   # no parameters.
  76   # accept input on stdin, put it on stdout.
  77   # patches accepted to get more stuff
  78   sed 's/\&quot;/\"/g
  79        s/\&amp;/\&/g
  80        s/\&lt;/</g
  81        s/\&gt;/>/g'
  82 }
  83
  84 ## basic S3 functionality.  x-amz-header functionality is not implemented.
  85 # make an S3 signature string, which will be output on stdout.
  86 s3_signature_string() {
  87   # $1 = HTTP verb
  88   # $2 = date string, must be in UTC
  89   # $3 = bucket name, if any
  90   # $4 = resource path, if any
  91   # $5 = content md5, if any
  92   # $6 = content MIME type, if any
  93   # $7 = canonicalized headers, if any
  94   # signature string will be output on stdout
  95   local verr="Must pass a verb to s3_signature_string!"
  96   local verb="${1:?verr}"
  97   local bucket="${3}"
  98   local resource="${4}"
  99   local derr="Must pass a date to s3_signature_string!"
 100   local date="${2:?derr}"
 101   local mime="${6}"
 102   local md5="${5}"
 103   local headers="${7}"
 104   printf "%s\n%s\n%s\n%s\n%s%s%s" \
 105     "${verb}" "${md5}" "${mime}" "${date}" \
 106     "${headers}" "${bucket}" "${resource}" | \
 107     $HMAC sha1 "${S3_SECRET_ACCESS_KEY}" | openssl base64 -e -a
 108 }
 109
 110 # cheesy, but it is the best way to have multiple headers.
 111 curl_headers() {
 112   # each arg passed will be output on its own line
 113   local parms=$#
 114   for ((;$#;)); do
 115     echo "header = \"${1}\""
 116     shift
 117   done
 118 }
 119
 120 s3_curl() {
 121   # invoke curl to do all the heavy HTTP lifting
 122   # $1 = method (one of GET, PUT, or DELETE. HEAD is not handled yet.)
 123   # $2 = remote bucket.
 124   # $3 = remote name
 125   # $4 = local name.
 126   # $5 = bandwidth limit.
 127   local bucket remote date sig md5 arg inout headers
 128   # header handling is kinda fugly, but it works.
 129   bucket="${2:+/${2}}/" # slashify the bucket
 130   remote="$(urlenc "${3}")" # if you don't, strange things may happen.
 131   stdopts="--connect-timeout 10 --fail --silent"
 132   [[ $CURL_S3_DEBUG == true ]] && stdopts="${stdopts} --show-error --fail"
 133   test -n "${5}" && stdopts="${stdopts} --limit-rate ${5}"
 134   case "${1}" in
 135    GET) arg="-o" inout="${4:--}" # stdout if no $4
 136         ;;
 137    PUT) [[ ${2} ]] || die "PUT can has bucket?"
 138         if [[ ! ${3} ]]; then
 139           arg="-X PUT"
 140           headers[${#headers[@]}]="Content-Length: 0"
 141         elif [[ -f ${4} ]]; then
 142           md5="$(openssl dgst -md5 -binary "${4}"|openssl base64 -e -a)"
 143           arg="-T" inout="${4}"
 144           headers[${#headers[@]}]="Expect: 100-continue"
 145         else
 146           die "Cannot write non-existing file ${4}"
 147         fi
 148         ;;
 149    DELETE) arg="-X DELETE"
 150            ;;
 151    HEAD) arg="-I" ;;
 152    *) die "Unknown verb ${1}.  It probably would not have worked anyways." ;;
 153   esac
 154   date="$(TZ=UTC date '+%a, %e %b %Y %H:%M:%S %z')"
 155   sig=$(s3_signature_string ${1} "${date}" "${bucket}" "${remote}" "${md5}")
 156
 157   headers[${#headers[@]}]="Authorization: AWS ${S3_ACCESS_KEY_ID}:${sig}"
 158   headers[${#headers[@]}]="Date: ${date}"
 159   [[ ${md5} ]] && headers[${#headers[@]}]="Content-MD5: ${md5}"
 160     $CURL ${arg} "${inout}" ${stdopts} -K <(curl_headers "${headers[@]}") \
 161        "http://s3.amazonaws.com${bucket}${remote}"
 162   return $?
 163 }
 164
 165 s3_put() {
 166   # $1 = remote bucket to put it into
 167   # $2 = remote name to put
 168   # $3 = file to put.  This must be present if $2 is.
 169   # $4 = bandwidth limit.
 170   s3_curl PUT "${1}" "${2}" "${3:-${2}}" "${4}"
 171   return $?
 172 }
 173
 174 s3_get() {
 175   # $1 = bucket to get file from
 176   # $2 = remote file to get
 177   # $3 = local file to get into. Will be overwritten if it exists.
 178   #      If this contains a path, that path must exist before calling this.
 179   # $4 = bandwidth limit.
 180   s3_curl GET "${1}" "${2}" "${3:-${2}}" "${4}"
 181   return $?
 182 }
 183
 184 s3_test() {
 185   # same args as s3_get, but uses the HEAD verb instead of the GET verb.
 186   s3_curl HEAD "${1}" "${2}" >/dev/null
 187   return $?
 188 }
 189
 190 # Hideously ugly, but it works well enough.
 191 s3_buckets() {
 192   s3_get |grep -o '<Name>[^>]*</Name>' |sed 's/<[^>]*>//g' |xmldec
 193   return $?
 194 }
 195
 196 # this will only return the first thousand entries, alas
 197 # Mabye some kind soul can fix this without writing an XML parser in bash?
 198 # Also need to add xml entity handling.
 199 s3_list() {
 200   # $1 = bucket to list
 201   [ "x${1}" == "x" ] && return 1
 202   s3_get "${1}" |grep -o '<Key>[^>]*</Key>' |sed 's/<[^>]*>//g'| xmldec
 203   return $?
 204 }
 205
 206 s3_delete() {
 207   # $1 = bucket to delete from
 208   # $2 = item to delete
 209   s3_curl DELETE "${1}" "${2}"
 210   return $?
 211 }
 212
 213 # because this uses s3_list, it suffers from the same flaws.
 214 s3_rmrf() {
 215   # $1 = bucket to delete everything from
 216   s3_list "${1}" | while read f; do
 217     s3_delete "${1}" "${f}";
 218   done
 219 }
 220
 221 check_deps
 222 case $1 in
 223   put) shift; s3_put "$@" ;;
 224   get) shift; s3_get "$@" ;;
 225   rm) shift; s3_delete "$@" ;;
 226   ls) shift; s3_list "$@" ;;
 227   test) shift; s3_test "$@" ;;
 228   buckets) s3_buckets ;;
 229   rmrf) shift; s3_rmrf "$@" ;;
 230   *) die "Unknown command ${1}."
 231     ;;
 232 esac
 233