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 ATTEMPTS=3
   9
  10 # print a message and bail
  11 die() {
  12   echo $*
  13   exit 1
  14 }
  15
  16 # check to see if the variable name passed exists and holds a value.
  17 # Die if it does not.
  18 check_or_die() {
  19   [[ ${!1} ]] || die "Environment variable ${1} is not set."
  20 }
  21
  22 # check to see if we have all the needed S3 variables defined.
  23 # Bail if we do not.
  24 check_s3() {
  25   local sak x
  26   for x in S3_ACCESS_KEY_ID S3_SECRET_ACCESS_KEY; do
  27     check_or_die ${x};
  28   done
  29   [[ -f ${S3_SECRET_ACCESS_KEY} ]] || die "S3_SECRET_ACCESS_KEY must point to a file!"
  30   sak="$(wc -c "${S3_SECRET_ACCESS_KEY}")"
  31   (( ${sak%%[!0-9 ]*} == 40 )) || \
  32     die "S3 Secret Access Key is not exactly 40 bytes long.  Please fix it."
  33 }
  34 # check to see if our external dependencies exist
  35 check_dep() {
  36   local res=0
  37   while [[ $# -ne 0 ]]; do
  38     which "${1}" >& /dev/null || { res=1; echo "${1} not found."; }
  39     shift
  40   done
  41   (( res == 0 )) || die "aborting."
  42 }
  43
  44 check_hmac() {
  45   if test ! -f $HMAC || test ! -x $HMAC; then
  46     die "hmac script not found or not executable."
  47   fi
  48 }
  49
  50 check_deps() {
  51   check_dep openssl date cat grep
  52   check_hmac
  53   check_s3
  54 }
  55
  56 urlenc() {
  57   # $1 = string to url encode
  58   # output is on stdout
  59   # we don't urlencode everything, just enough stuff.
  60   echo -n "${1}" |
  61   sed 's/%/%25/g
  62        s/ /%20/g
  63        s/#/%23/g
  64        s/\$/%24/g
  65        s/\&/%26/g
  66        s/+/%2b/g
  67        s/,/%2c/g
  68        s/:/%3a/g
  69        s/;/%3b/g
  70        s/?/%3f/g
  71        s/@/%40/g
  72        s/       /%09/g'
  73 }
  74
  75 xmldec() {
  76   # no parameters.
  77   # accept input on stdin, put it on stdout.
  78   # patches accepted to get more stuff
  79   sed 's/\&quot;/\"/g
  80        s/\&amp;/\&/g
  81        s/\&lt;/</g
  82        s/\&gt;/>/g'
  83 }
  84
  85 ## basic S3 functionality.  x-amz-header functionality is not implemented.
  86 # make an S3 signature string, which will be output on stdout.
  87 s3_signature_string() {
  88   # $1 = HTTP verb
  89   # $2 = date string, must be in UTC
  90   # $3 = bucket name, if any
  91   # $4 = resource path, if any
  92   # $5 = content md5, if any
  93   # $6 = content MIME type, if any
  94   # $7 = canonicalized headers, if any
  95   # signature string will be output on stdout
  96   local verr="Must pass a verb to s3_signature_string!"
  97   local verb="${1:?verr}"
  98   local bucket="${3}"
  99   local resource="${4}"
 100   local derr="Must pass a date to s3_signature_string!"
 101   local date="${2:?derr}"
 102   local mime="${6}"
 103   local md5="${5}"
 104   local headers="${7}"
 105   printf "%s\n%s\n%s\n%s\n%s%s%s" \
 106     "${verb}" "${md5}" "${mime}" "${date}" \
 107     "${headers}" "${bucket}" "${resource}" | \
 108     $HMAC sha1 "${S3_SECRET_ACCESS_KEY}" | openssl base64 -e -a
 109 }
 110
 111 # cheesy, but it is the best way to have multiple headers.
 112 curl_headers() {
 113   # each arg passed will be output on its own line
 114   local parms=$#
 115   for ((;$#;)); do
 116     echo "header = \"${1}\""
 117     shift
 118   done
 119 }
 120
 121 s3_curl() {
 122   # invoke curl to do all the heavy HTTP lifting
 123   # $1 = method (one of GET, PUT, or DELETE. HEAD is not handled yet.)
 124   # $2 = remote bucket.
 125   # $3 = remote name
 126   # $4 = local name.
 127   # $5 = bandwidth limit.
 128   local bucket remote date sig md5 arg inout headers tries ret
 129   # header handling is kinda fugly, but it works.
 130   bucket="${2:+/${2}}/" # slashify the bucket
 131   remote="$(urlenc "${3}")" # if you don't, strange things may happen.
 132   stdopts="--connect-timeout 10 --fail --silent"
 133   [[ $CURL_S3_DEBUG == true ]] && stdopts="${stdopts} --show-error --fail"
 134   test -n "${5}" && stdopts="${stdopts} --limit-rate ${5}"
 135   case "${1}" in
 136    GET) arg="-o" inout="${4:--}" # stdout if no $4
 137         ;;
 138    PUT) [[ ${2} ]] || die "PUT can has bucket?"
 139         if [[ ! ${3} ]]; then
 140           arg="-X PUT"
 141           headers[${#headers[@]}]="Content-Length: 0"
 142         elif [[ -f ${4} ]]; then
 143           md5="$(openssl dgst -md5 -binary "${4}"|openssl base64 -e -a)"
 144           arg="-T" inout="${4}"
 145           headers[${#headers[@]}]="Expect: 100-continue"
 146         else
 147           die "Cannot write non-existing file ${4}"
 148         fi
 149         ;;
 150    DELETE) arg="-X DELETE"
 151            ;;
 152    HEAD) arg="-I" ;;
 153    *) die "Unknown verb ${1}.  It probably would not have worked anyways." ;;
 154   esac
 155   date="$(TZ=UTC date '+%a, %e %b %Y %H:%M:%S %z')"
 156   sig=$(s3_signature_string ${1} "${date}" "${bucket}" "${remote}" "${md5}")
 157
 158   headers[${#headers[@]}]="Authorization: AWS ${S3_ACCESS_KEY_ID}:${sig}"
 159   headers[${#headers[@]}]="Date: ${date}"
 160   [[ ${md5} ]] && headers[${#headers[@]}]="Content-MD5: ${md5}"
 161   tries=0
 162   while test $tries -lt 3; do
 163     tries=$(expr $tries + 1)
 164     $CURL ${arg} "${inout}" ${stdopts} -K <(curl_headers "${headers[@]}") \
 165        "http://s3.amazonaws.com${bucket}${remote}"
 166     ret=$?
 167     test $ret -eq 0 && break;
 168   done
 169   return $ret
 170 }
 171
 172 s3_put() {
 173   # $1 = remote bucket to put it into
 174   # $2 = remote name to put
 175   # $3 = file to put.  This must be present if $2 is.
 176   # $4 = bandwidth limit.
 177   s3_curl PUT "${1}" "${2}" "${3:-${2}}" "${4}"
 178   return $?
 179 }
 180
 181 s3_get() {
 182   # $1 = bucket to get file from
 183   # $2 = remote file to get
 184   # $3 = local file to get into. Will be overwritten if it exists.
 185   #      If this contains a path, that path must exist before calling this.
 186   # $4 = bandwidth limit.
 187   s3_curl GET "${1}" "${2}" "${3:-${2}}" "${4}"
 188   return $?
 189 }
 190
 191 s3_test() {
 192   # same args as s3_get, but uses the HEAD verb instead of the GET verb.
 193   s3_curl HEAD "${1}" "${2}" >/dev/null
 194   return $?
 195 }
 196
 197 # Hideously ugly, but it works well enough.
 198 s3_buckets() {
 199   s3_get |grep -o '<Name>[^>]*</Name>' |sed 's/<[^>]*>//g' |xmldec
 200   return $?
 201 }
 202
 203 # this will only return the first thousand entries, alas
 204 # Mabye some kind soul can fix this without writing an XML parser in bash?
 205 # Also need to add xml entity handling.
 206 s3_list() {
 207   # $1 = bucket to list
 208   [ "x${1}" == "x" ] && return 1
 209   s3_get "${1}" |grep -o '<Key>[^>]*</Key>' |sed 's/<[^>]*>//g'| xmldec
 210   return $?
 211 }
 212
 213 s3_delete() {
 214   # $1 = bucket to delete from
 215   # $2 = item to delete
 216   s3_curl DELETE "${1}" "${2}"
 217   return $?
 218 }
 219
 220 # because this uses s3_list, it suffers from the same flaws.
 221 s3_rmrf() {
 222   # $1 = bucket to delete everything from
 223   s3_list "${1}" | while read f; do
 224     s3_delete "${1}" "${f}";
 225   done
 226 }
 227
 228 check_deps
 229 case $1 in
 230   put) shift; s3_put "$@" ;;
 231   get) shift; s3_get "$@" ;;
 232   rm) shift; s3_delete "$@" ;;
 233   ls) shift; s3_list "$@" ;;
 234   test) shift; s3_test "$@" ;;
 235   buckets) s3_buckets ;;
 236   rmrf) shift; s3_rmrf "$@" ;;
 237   *) die "Unknown command ${1}."
 238     ;;
 239 esac
 240