auth: add basic authentication filter framework

This leverages the new lua support. See
filters/simple-authentication.lua for explaination of how this works.
There is also additional documentation in cgitrc.5.txt.

Though this is a cookie-based approach, cgit's caching mechanism is
preserved for authenticated pages.

Very plugable and extendable depending on user needs.

The sample script uses an HMAC-SHA1 based cookie to store the
currently logged in user, with an expiration date.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

6 files changed, 387 insertions, 16 deletions

diff --git a/cgit.c b/cgit.c
index f3fe56b..c52ef33 100644
--- a/cgit.c
+++ b/cgit.c
@@ -192,6 +192,8 @@ static void config_cb(const char *name, const char *value)
 		ctx.cfg.commit_filter = cgit_new_filter(value, COMMIT);
 	else if (!strcmp(name, "email-filter"))
 		ctx.cfg.email_filter = cgit_new_filter(value, EMAIL);
+	else if (!strcmp(name, "auth-filter"))
+		ctx.cfg.auth_filter = cgit_new_filter(value, AUTH);
 	else if (!strcmp(name, "embedded"))
 		ctx.cfg.embedded = atoi(value);
 	else if (!strcmp(name, "max-atom-items"))
@@ -378,6 +380,10 @@ static void prepare_context(struct cgit_context *ctx)
 	ctx->env.script_name = getenv("SCRIPT_NAME");
 	ctx->env.server_name = getenv("SERVER_NAME");
 	ctx->env.server_port = getenv("SERVER_PORT");
+	ctx->env.http_cookie = getenv("HTTP_COOKIE");
+	ctx->env.http_referer = getenv("HTTP_REFERER");
+	ctx->env.content_length = getenv("CONTENT_LENGTH") ? strtoul(getenv("CONTENT_LENGTH"), NULL, 10) : 0;
+	ctx->env.authenticated = 0;
 	ctx->page.mimetype = "text/html";
 	ctx->page.charset = PAGE_ENCODING;
 	ctx->page.filename = NULL;
@@ -593,11 +599,92 @@ static int prepare_repo_cmd(struct cgit_context *ctx)
 	return 0;
 }
 
+static inline void open_auth_filter(struct cgit_context *ctx, const char *function)
+{
+	cgit_open_filter(ctx->cfg.auth_filter, function,
+		ctx->env.http_cookie ? ctx->env.http_cookie : "",
+		ctx->env.request_method ? ctx->env.request_method : "",
+		ctx->env.query_string ? ctx->env.query_string : "",
+		ctx->env.http_referer ? ctx->env.http_referer : "",
+		ctx->env.path_info ? ctx->env.path_info : "",
+		ctx->env.http_host ? ctx->env.http_host : "",
+		ctx->env.https ? ctx->env.https : "",
+		ctx->qry.repo ? ctx->qry.repo : "",
+		ctx->qry.page ? ctx->qry.page : "",
+		ctx->qry.url ? ctx->qry.url : "");
+}
+
+#define MAX_AUTHENTICATION_POST_BYTES 4096
+static inline void authenticate_post(struct cgit_context *ctx)
+{
+	if (ctx->env.http_referer && strlen(ctx->env.http_referer) > 0) {
+		html("Status: 302 Redirect\n");
+		html("Cache-Control: no-cache, no-store\n");
+		htmlf("Location: %s\n", ctx->env.http_referer);
+	} else {
+		html("Status: 501 Missing Referer\n");
+		html("Cache-Control: no-cache, no-store\n\n");
+		exit(0);
+	}
+
+	open_auth_filter(ctx, "authenticate-post");
+	char buffer[MAX_AUTHENTICATION_POST_BYTES];
+	int len;
+	len = ctx->env.content_length;
+	if (len > MAX_AUTHENTICATION_POST_BYTES)
+		len = MAX_AUTHENTICATION_POST_BYTES;
+	if (read(STDIN_FILENO, buffer, len) < 0)
+		die_errno("Could not read POST from stdin");
+	if (write(STDOUT_FILENO, buffer, len) < 0)
+		die_errno("Could not write POST to stdout");
+	/* The filter may now spit out a Set-Cookie: ... */
+	cgit_close_filter(ctx->cfg.auth_filter);
+
+	html("\n");
+	exit(0);
+}
+
+static inline void authenticate_cookie(struct cgit_context *ctx)
+{
+	/* If we don't have an auth_filter, consider all cookies valid, and thus return early. */
+	if (!ctx->cfg.auth_filter) {
+		ctx->env.authenticated = 1;
+		return;
+	}
+
+	/* If we're having something POST'd to /login, we're authenticating POST,
+	 * instead of the cookie, so call authenticate_post and bail out early.
+	 * This pattern here should match /?p=login with POST. */
+	if (ctx->env.request_method && ctx->qry.page && !ctx->repo && \
+	    !strcmp(ctx->env.request_method, "POST") && !strcmp(ctx->qry.page, "login")) {
+		authenticate_post(ctx);
+		return;
+	}
+
+	/* If we've made it this far, we're authenticating the cookie for real, so do that. */
+	open_auth_filter(ctx, "authenticate-cookie");
+	ctx->env.authenticated = cgit_close_filter(ctx->cfg.auth_filter);
+}
+
 static void process_request(void *cbdata)
 {
 	struct cgit_context *ctx = cbdata;
 	struct cgit_cmd *cmd;
 
+	/* If we're not yet authenticated, no matter what page we're on,
+	 * display the authentication body from the auth_filter. This should
+	 * never be cached. */
+	if (!ctx->env.authenticated) {
+		ctx->page.title = "Authentication Required";
+		cgit_print_http_headers(ctx);
+		cgit_print_docstart(ctx);
+		cgit_print_pageheader(ctx);
+		open_auth_filter(ctx, "body");
+		cgit_close_filter(ctx->cfg.auth_filter);
+		cgit_print_docend();
+		return;
+	}
+
 	cmd = cgit_get_cmd(ctx);
 	if (!cmd) {
 		ctx->page.title = "cgit error";
@@ -911,6 +998,7 @@ int main(int argc, const char **argv)
 	int err, ttl;
 
 	cgit_init_filters();
+	atexit(cgit_cleanup_filters);
 
 	prepare_context(&ctx);
 	cgit_repolist.length = 0;
@@ -948,18 +1036,22 @@ int main(int argc, const char **argv)
 		cgit_parse_url(ctx.qry.url);
 	}
 
+	/* Before we go any further, we set ctx.env.authenticated by checking to see
+	 * if the supplied cookie is valid. All cookies are valid if there is no
+	 * auth_filter. If there is an auth_filter, the filter decides. */
+	authenticate_cookie(&ctx);
+
 	ttl = calc_ttl();
 	if (ttl < 0)
 		ctx.page.expires += 10 * 365 * 24 * 60 * 60; /* 10 years */
 	else
 		ctx.page.expires += ttl * 60;
-	if (ctx.env.request_method && !strcmp(ctx.env.request_method, "HEAD"))
+	if (!ctx.env.authenticated || (ctx.env.request_method && !strcmp(ctx.env.request_method, "HEAD")))
 		ctx.cfg.nocache = 1;
 	if (ctx.cfg.nocache)
 		ctx.cfg.cache_size = 0;
 	err = cache_process(ctx.cfg.cache_size, ctx.cfg.cache_root,
 			    ctx.qry.raw, ttl, process_request, &ctx);
-	cgit_cleanup_filters();
 	if (err)
 		cgit_print_error("Error processing page: %s (%d)",
 				 strerror(err), err);
diff --git a/cgit.h b/cgit.h
index e200a06..496d0f6 100644
--- a/cgit.h
+++ b/cgit.h
@@ -53,7 +53,7 @@ typedef void (*filepair_fn)(struct diff_filepair *pair);
 typedef void (*linediff_fn)(char *line, int len);
 
 typedef enum {
-	ABOUT, COMMIT, SOURCE, EMAIL
+	ABOUT, COMMIT, SOURCE, EMAIL, AUTH
 } filter_type;
 
 struct cgit_filter {
@@ -252,6 +252,7 @@ struct cgit_config {
 	struct cgit_filter *commit_filter;
 	struct cgit_filter *source_filter;
 	struct cgit_filter *email_filter;
+	struct cgit_filter *auth_filter;
 };
 
 struct cgit_page {
@@ -278,6 +279,10 @@ struct cgit_environment {
 	const char *script_name;
 	const char *server_name;
 	const char *server_port;
+	const char *http_cookie;
+	const char *http_referer;
+	unsigned int content_length;
+	int authenticated;
 };
 
 struct cgit_context {
diff --git a/cgitrc.5.txt b/cgitrc.5.txt
index 170e825..c45dbd3 100644
--- a/cgitrc.5.txt
+++ b/cgitrc.5.txt
@@ -42,6 +42,13 @@ agefile::
 	hh:mm:ss". You may want to generate this file from a post-receive
 	hook. Default value: "info/web/last-modified".
 
+auth-filter::
+	Specifies a command that will be invoked for authenticating repository
+	access. Receives quite a few arguments, and data on both stdin and
+	stdout for authentication processing. Details follow later in this
+	document. If no auth-filter is specified, no authentication is
+	performed. Default value: none. See also: "FILTER API".
+
 branch-sort::
 	Flag which, when set to "age", enables date ordering in the branch ref
 	list, and when set to "name" enables ordering by branch name. Default
@@ -605,6 +612,8 @@ specification with the relevant string; available values are:
 		URL escapes for a path and writes 'str' to the webpage.
 	'html_url_arg(str)'::
 		URL escapes for an argument and writes 'str' to the webpage.
+	'html_include(file)'::
+		Includes 'file' in webpage.
 
 
 Parameters are provided to filters as follows.
@@ -635,7 +644,32 @@ source filter::
 	file that is to be filtered is available on standard input and the
 	filtered contents is expected on standard output.
 
-Also, all filters are handed the following environment variables:
+auth filter::
+	The authentication filter receives 11 parameters:
+	  - filter action, explained below, which specifies which action the
+	    filter is called for
+	  - http cookie
+	  - http method
+	  - http referer
+	  - http path
+	  - http https flag
+	  - cgit repo
+	  - cgit page
+	  - cgit url
+	When the filter action is "body", this filter must write to output the
+	HTML for displaying the login form, which POSTs to "/?p=login". When
+	the filter action is "authenticate-cookie", this filter must validate
+	the http cookie and return a 0 if it is invalid or 1 if it is invalid,
+	in the exit code / close function. If the filter action is
+	"authenticate-post", this filter receives POST'd parameters on
+	standard input, and should write to output one or more "Set-Cookie"
+	HTTP headers, each followed by a newline.
+
+	Please see `filters/simple-authentication.lua` for a clear example
+	script that may be modified.
+
+
+All filters are handed the following environment variables:
 
 - CGIT_REPO_URL (from repo.url)
 - CGIT_REPO_NAME (from repo.name)
diff --git a/filter.c b/filter.c
index 0cce7bb..a5e5e4b 100644
--- a/filter.c
+++ b/filter.c
@@ -244,6 +244,11 @@ static int html_url_arg_lua_filter(lua_State *lua_state)
 	return hook_lua_filter(lua_state, html_url_arg);
 }
 
+static int html_include_lua_filter(lua_State *lua_state)
+{
+	return hook_lua_filter(lua_state, (void (*)(const char *))html_include);
+}
+
 static void cleanup_lua_filter(struct cgit_filter *base)
 {
 	struct lua_filter *filter = (struct lua_filter *)base;
@@ -279,6 +284,8 @@ static int init_lua_filter(struct lua_filter *filter)
 	lua_setglobal(filter->lua_state, "html_url_path");
 	lua_pushcfunction(filter->lua_state, html_url_arg_lua_filter);
 	lua_setglobal(filter->lua_state, "html_url_arg");
+	lua_pushcfunction(filter->lua_state, html_include_lua_filter);
+	lua_setglobal(filter->lua_state, "html_include");
 
 	if (luaL_dofile(filter->lua_state, filter->script_file)) {
 		error_lua_filter(filter);
@@ -409,6 +416,10 @@ struct cgit_filter *cgit_new_filter(const char *cmd, filter_type filtertype)
 		colon = NULL;
 
 	switch (filtertype) {
+		case AUTH:
+			argument_count = 11;
+			break;
+
 		case EMAIL:
 			argument_count = 2;
 			break;
diff --git a/filters/simple-authentication.lua b/filters/simple-authentication.lua
new file mode 100644
index 0000000..4cd4983
--- /dev/null
+++ b/filters/simple-authentication.lua
@@ -0,0 +1,225 @@
+-- This script may be used with the auth-filter. Be sure to configure it as you wish.
+--
+-- Requirements:
+-- 	luacrypto >= 0.3
+-- 	<http://mkottman.github.io/luacrypto/>
+--
+
+
+--
+--
+-- Configure these variables for your settings.
+--
+--
+
+local protected_repos = {
+	glouglou	= { laurent = true, jason = true },
+	qt		= { jason = true, bob = true }
+}
+
+local users = {
+	jason		= "secretpassword",
+	laurent		= "s3cr3t",
+	bob		= "ilikelua"
+}
+
+local secret = "BE SURE TO CUSTOMIZE THIS STRING TO SOMETHING BIG AND RANDOM"
+
+
+
+--
+--
+-- Authentication functions follow below. Swap these out if you want different authentication semantics.
+--
+--
+
+-- Sets HTTP cookie headers based on post
+function authenticate_post()
+	local password = users[post["username"]]
+	-- TODO: Implement time invariant string comparison function to mitigate against timing attack.
+	if password == nil or password ~= post["password"] then
+		construct_cookie("", "cgitauth")
+	else
+		construct_cookie(post["username"], "cgitauth")
+	end
+	return 0
+end
+
+
+-- Returns 1 if the cookie is valid and 0 if it is not.
+function authenticate_cookie()
+	accepted_users = protected_repos[cgit["repo"]]
+	if accepted_users == nil then
+		-- We return as valid if the repo is not protected.
+		return 1
+	end
+
+	local username = validate_cookie(get_cookie(http["cookie"], "cgitauth"))
+	if username == nil or not accepted_users[username] then
+		return 0
+	else
+		return 1
+	end
+end
+
+-- Prints the html for the login form.
+function body()
+	html("<h2>Authentication Required</h2>")
+	html("<form method='post' action='")
+	html_attr(cgit["login"])
+	html("'>")
+	html("<table>")
+	html("<tr><td><label for='username'>Username:</label></td><td><input id='username' name='username' autofocus /></td></tr>")
+	html("<tr><td><label for='password'>Password:</label></td><td><input id='password' name='password' type='password' /></td></tr>")
+	html("<tr><td colspan='2'><input value='Login' type='submit' /></td></tr>")
+	html("</table></form>")
+
+	return 0
+end
+
+
+--
+--
+-- Cookie construction and validation helpers.
+--
+--
+
+local crypto = require("crypto")
+
+-- Returns username of cookie if cookie is valid. Otherwise returns nil.
+function validate_cookie(cookie)
+	local i = 0
+	local username = ""
+	local expiration = 0
+	local salt = ""
+	local hmac = ""
+
+	if cookie:len() < 3 or cookie:sub(1, 1) == "|" then
+		return nil
+	end
+
+	for component in string.gmatch(cookie, "[^|]+") do
+		if i == 0 then
+			username = component
+		elseif i == 1 then
+			expiration = tonumber(component)
+			if expiration == nil then
+				expiration = 0
+			end
+		elseif i == 2 then
+			salt = component
+		elseif i == 3 then
+			hmac = component
+		else
+			break
+		end
+		i = i + 1
+	end
+
+	if hmac == nil or hmac:len() == 0 then
+		return nil
+	end
+
+	-- TODO: implement time invariant comparison to prevent against timing attack.
+	if hmac ~= crypto.hmac.digest("sha1", username .. "|" .. tostring(expiration) .. "|" .. salt, secret) then
+		return nil
+	end
+
+	if expiration <= os.time() then
+		return nil
+	end
+
+	return username:lower()
+end
+
+function construct_cookie(username, cookie)
+	local authstr = ""
+	if username:len() > 0 then
+		-- One week expiration time
+		local expiration = os.time() + 604800
+		local salt = crypto.hex(crypto.rand.bytes(16))
+
+		authstr = username .. "|" .. tostring(expiration) .. "|" .. salt
+		authstr = authstr .. "|" .. crypto.hmac.digest("sha1", authstr, secret)
+	end
+
+	html("Set-Cookie: " .. cookie .. "=" .. authstr .. "; HttpOnly")
+	if http["https"] == "yes" or http["https"] == "on" or http["https"] == "1" then
+		html("; secure")
+	end
+	html("\n")
+end
+
+--
+--
+-- Wrapper around filter API follows below, exposing the http table, the cgit table, and the post table to the above functions.
+--
+--
+
+local actions = {}
+actions["authenticate-post"] = authenticate_post
+actions["authenticate-cookie"] = authenticate_cookie
+actions["body"] = body
+
+function filter_open(...)
+	action = actions[select(1, ...)]
+
+	http = {}
+	http["cookie"] = select(2, ...)
+	http["method"] = select(3, ...)
+	http["query"] = select(4, ...)
+	http["referer"] = select(5, ...)
+	http["path"] = select(6, ...)
+	http["host"] = select(7, ...)
+	http["https"] = select(8, ...)
+
+	cgit = {}
+	cgit["repo"] = select(9, ...)
+	cgit["page"] = select(10, ...)
+	cgit["url"] = select(11, ...)
+
+	cgit["login"] = ""
+	for _ in cgit["url"]:gfind("/") do
+		cgit["login"] = cgit["login"] .. "../"
+	end
+	cgit["login"] = cgit["login"] .. "?p=login"
+
+end
+
+function filter_close()
+	return action()
+end
+
+function filter_write(str)
+	post = parse_qs(str)
+end
+
+
+--
+--
+-- Utility functions follow below, based on keplerproject/wsapi.
+--
+--
+
+function url_decode(str)
+	if not str then
+		return ""
+	end
+	str = string.gsub(str, "+", " ")
+	str = string.gsub(str, "%%(%x%x)", function(h) return string.char(tonumber(h, 16)) end)
+	str = string.gsub(str, "\r\n", "\n")
+	return str
+end
+
+function parse_qs(qs)
+	local tab = {}
+	for key, val in string.gmatch(qs, "([^&=]+)=([^&=]*)&?") do
+		tab[url_decode(key)] = url_decode(val)
+	end
+	return tab
+end
+
+function get_cookie(cookies, name)
+	cookies = string.gsub(";" .. cookies .. ";", "%s*;%s*", ";")
+	return url_decode(string.match(cookies, ";" .. name .. "=(.-);"))
+end
diff --git a/ui-shared.c b/ui-shared.c
index abe15cd..4f47c50 100644
--- a/ui-shared.c
+++ b/ui-shared.c
@@ -641,6 +641,8 @@ void cgit_print_http_headers(struct cgit_context *ctx)
 	if (ctx->page.filename)
 		htmlf("Content-Disposition: inline; filename=\"%s\"\n",
 		      ctx->page.filename);
+	if (!ctx->env.authenticated)
+		html("Cache-Control: no-cache, no-store\n");
 	htmlf("Last-Modified: %s\n", http_date(ctx->page.modified));
 	htmlf("Expires: %s\n", http_date(ctx->page.expires));
 	if (ctx->page.etag)
@@ -814,14 +816,16 @@ static void print_header(struct cgit_context *ctx)
 		cgit_index_link("index", NULL, NULL, NULL, NULL, 0);
 		html(" : ");
 		cgit_summary_link(ctx->repo->name, ctx->repo->name, NULL, NULL);
-		html("</td><td class='form'>");
-		html("<form method='get' action=''>\n");
-		cgit_add_hidden_formfields(0, 1, ctx->qry.page);
-		html("<select name='h' onchange='this.form.submit();'>\n");
-		for_each_branch_ref(print_branch_option, ctx->qry.head);
-		html("</select> ");
-		html("<input type='submit' name='' value='switch'/>");
-		html("</form>");
+		if (ctx->env.authenticated) {
+			html("</td><td class='form'>");
+			html("<form method='get' action=''>\n");
+			cgit_add_hidden_formfields(0, 1, ctx->qry.page);
+			html("<select name='h' onchange='this.form.submit();'>\n");
+			for_each_branch_ref(print_branch_option, ctx->qry.head);
+			html("</select> ");
+			html("<input type='submit' name='' value='switch'/>");
+			html("</form>");
+		}
 	} else
 		html_txt(ctx->cfg.root_title);
 	html("</td></tr>\n");
@@ -843,11 +847,11 @@ static void print_header(struct cgit_context *ctx)
 void cgit_print_pageheader(struct cgit_context *ctx)
 {
 	html("<div id='cgit'>");
-	if (!ctx->cfg.noheader)
+	if (!ctx->env.authenticated || !ctx->cfg.noheader)
 		print_header(ctx);
 
 	html("<table class='tabs'><tr><td>\n");
-	if (ctx->repo) {
+	if (ctx->env.authenticated && ctx->repo) {
 		cgit_summary_link("summary", NULL, hc(ctx, "summary"),
 				  ctx->qry.head);
 		cgit_refs_link("refs", NULL, hc(ctx, "refs"), ctx->qry.head,
@@ -886,7 +890,7 @@ void cgit_print_pageheader(struct cgit_context *ctx)
 		html("'/>\n");
 		html("<input type='submit' value='search'/>\n");
 		html("</form>\n");
-	} else {
+	} else if (ctx->env.authenticated) {
 		site_link(NULL, "index", NULL, hc(ctx, "repolist"), NULL, NULL, 0);
 		if (ctx->cfg.root_readme)
 			site_link("about", "about", NULL, hc(ctx, "about"),
@@ -902,7 +906,7 @@ void cgit_print_pageheader(struct cgit_context *ctx)
 		html("</form>");
 	}
 	html("</td></tr></table>\n");
-	if (ctx->qry.vpath) {
+	if (ctx->env.authenticated && ctx->qry.vpath) {
 		html("<div class='path'>");
 		html("path: ");
 		cgit_print_path_crumbs(ctx, ctx->qry.vpath);