git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 5/5] git-daemon support for user-relative paths.

From
Junio C Hamano <junkio@cox.net>
Date
Nov 21, 2005, 09:28 UTC
Message-ID
<7vfypquz88.fsf@assigned-by-dhcp.cox.net>
In-Reply-To
<437E67AC.2010400@op5.se>
Andreas Ericsson <ae@op5.se> writes:
> I'll run the clone/fetch/push test-suite again tomorrow, with this 
> applied. It looks good though.

Sorry, but there was a thinko in my butchered version of enter_repo(). While allowing only absolute path was good for the version with your daemon.c change, it was not with the current one that runs upload-pack with "." as repo. In either case we _do_ chdir() to it after validating the path, so I am wondering if it is a good idea to keep sending "." as repo when executing upload-pack with this patch as well. This does not make any practical difference, but I think it makes the intent clearer -- "we are already there so do not try going anywhere else".

So I am thinking about applying something like this patch on top of the last part of your patch.

 - Do validation only on canonicalized paths;
 - Run upload-pack with "." as repo, not full path;
 - allow trailing slash under --strict-paths i.e. "git://host/my/repo.git/"
What do you think?
---
diff --git a/daemon.c b/daemon.c
index ac4c94b..21d8260 100644
--- a/daemon.c
+++ b/daemon.c
@@ -93,22 +93,19 @@ static char *path_ok(char *dir)
 
 	if ( ok_paths && *ok_paths ) {
 		char **pp = NULL;
-		int dirlen = strlen(dir);
 		int pathlen = strlen(path);
 
+		/* The validation is done on the paths after enter_repo
+		 * canonicalization, so whitelist should be written in
+		 * terms of real pathnames (i.e. after ~user is expanded
+		 * and symlinks resolved).
+		 */
 		for ( pp = ok_paths ; *pp ; pp++ ) {
 			int len = strlen(*pp);
-			/* because of symlinks we must match both what the
-			 * user passed and the canonicalized path, otherwise
-			 * the user can send a string matching either a whitelist
-			 * entry or an actual directory exactly and still not
-			 * get through */
 			if (len <= pathlen && !memcmp(*pp, path, len)) {
-				if (path[len] == '\0' || (!strict_paths && path[len] == '/'))
-					return path;
-			}
-			if (len <= dirlen && !memcmp(*pp, dir, len)) {
-				if (dir[len] == '\0' || (!strict_paths && dir[len] == '/'))
+				if (path[len] == '\0' ||
+				    (path[len] == '/' &&
+				     (!strict_paths || path[len+1] == 0)))
 					return path;
 			}
 		}
@@ -160,7 +157,7 @@ static int upload(char *dir)
 	snprintf(timeout_buf, sizeof timeout_buf, "--timeout=%u", timeout);
 
 	/* git-upload-pack only ever reads stuff, so this is safe */
-	execlp("git-upload-pack", "git-upload-pack", "--strict", timeout_buf, path, NULL);
+	execlp("git-upload-pack", "git-upload-pack", "--strict", timeout_buf, ".", NULL);
 	return -1;
 }
 


 
Previous: Andreas EricssonNext: Junio C Hamano
Message 9 of 12 in “git-daemon support for user-relative paths.”
  1. 5/5 git-daemon support for user-relative paths.Andreas Ericsson, Nov 17, 2005
  2. Junio C HamanoNov 18, 2005
  3. Andreas EricssonNov 18, 2005
  4. Matthias UrlichsNov 18, 2005
  5. H. Peter AnvinNov 18, 2005
  6. Junio C HamanoNov 18, 2005
  7. Junio C HamanoNov 18, 2005
  8. Andreas EricssonNov 18, 2005
  9. Junio C HamanoNov 21, 2005
  10. Junio C HamanoNov 21, 2005
  11. Andreas EricssonNov 21, 2005
  12. Junio C HamanoNov 21, 2005

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.