Cgi 工具 |
|
print"Content-type: text/html; charset=iso-8859-1\r\n\r\n"
-- Written by RiciLake. -- The author places the code into the public domain, renouncing all rights and responsibilities. -- Don't use this in a real application, see notes below. function parsecgi(str) local rv = {} for pair in str:gmatch"[^&]+" do local key, val = pair:match"([^=]*)=(.*)" if key then rv[key] = val end end return rv end
此实现存在多种安全隐患
=
+
或 %xx
进行解码(虽然在我看来,有效的键不需要 % 编码)
所以这里有一个稍微好一点的版本,它仍然没有处理“搜索”情况。您必须提供一个新创建的合法键表;如果一个键可以有多个值,您提供一个表作为关联值(它会被填充),否则您提供默认值(或 false
)。非法键会被忽略或抛出错误。没有 =
的段落会抛出错误。抛出错误是因为无效的查询/发布字符串很可能是攻击企图,并且应该(在我看来)被拒绝;因此,您应该将 CGI 处理程序包装在 pcall
中,并向浏览器返回 403
或 404
错误。
该函数故意不尝试对键进行 %-解码,因为有效的键永远不需要 %-编码,因此这种形式的 URL 很可能是一种模糊的攻击服务器的企图。
一个可能的改进是,如果初始表中默认值为数字,则还检查提供的 value 是否为数字。
-- Written by RiciLake. -- The author places the code into the public domain, renouncing all rights and responsibilities. -- Replace + with space and %xx with the corresponding character. local function cgidecode(str) return (str:gsub('+', ' '):gsub("%%(%x%x)", function(xx) return string.char(tonumber(xx, 16)) end)) end -- Main function -- Sample invocation: cgivals = parsecgi(str, {count = 10, start = 1, names = {}}) function parsecgi(str, keys, ignore_invalid) local keyfound = {} for pair in str:gmatch"[^&]+" do local key, val = pair:match"([^=]*)=(.*)" if not key then error"Invalid query string" end local default = keys[key] if default == nil then if not ignore_invalid then error"Invalid query string" end else if type(default) == "table" then default[#default+1] = cgidecode(val) elseif keyfound[key] then error"Invalid query string" else keyfound[key] = true keys[key] = cgidecode(val) end end end return keys end
str = os.getenv("QUERY_STRING") ...
-- Assigns header information to variable "l", and returns "Hello!" back as the webpage content. -- This could use some cleaning up - error checking, removal of repetition, improved scoping. -- The SCGI protocol has the HTTP requests forwarded to a specified port (default 4000), with -- headers passed directly through TCP (and an ASCII string length prefix followed by a ":", -- with the key-value pairs of the header seperated by null characters "\0") -- With no parsing, this code handles about 950 requests per second on a 2 year old laptop. local socket = require("socket") local host = host or "*" local port = port or 4000 local s = assert(socket.bind(host, port)) local i, p = s:getsockname() assert(i, p) print("Waiting on " .. i .. ":" .. p .. "...") while 1 do c = assert(s:accept()) print("Connection requested.") len = "" l, e = c:receive(1) while not e do if l == ":" then header_len = tonumber(len) ; break end len = len .. l l, e = c:receive(1) end l,e = c:receive(header_len) c:send("Status: 200 OK\r\n") c:send("Content-Type: text/plain\r\n") c:send("\r\n") c:send("Hello!") c:close() end
以下 httpd.conf (Apache 2) 中的代码行将允许此 SCGI 示例正常工作 - 尽管 127.0.0.1 需要更改为更合适的 IP 地址。
LoadModule scgi_module modules/mod_scgi.so SCGIMount /dynamic 127.0.0.1:4000