]> git.ipfire.org Git - thirdparty/git.git/commitdiff
submodule-config.h: move check_submodule_url
authorVictoria Dye <vdye@github.com>
Thu, 18 Jan 2024 01:55:15 +0000 (01:55 +0000)
committerJunio C Hamano <gitster@pobox.com>
Thu, 18 Jan 2024 18:12:48 +0000 (10:12 -0800)
Move 'check_submodule_url' out of 'fsck.c' and into 'submodule-config.h' as
a public method, similar to 'check_submodule_name'. With the function now
accessible outside of 'fsck', it can be used in a later commit to extend
'test-tool submodule' to check the validity of submodule URLs as it does
with names in the 'check-name' subcommand.

Other than its location, no changes are made to 'check_submodule_url' in
this patch.

Signed-off-by: Victoria Dye <vdye@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
fsck.c
submodule-config.c
submodule-config.h

diff --git a/fsck.c b/fsck.c
index 6a0bbc50877710ff22db33adec48f4a469143c37..129bc0630d34a3ea65f25cb9287bfb0d25f47fc4 100644 (file)
--- a/fsck.c
+++ b/fsck.c
@@ -21,7 +21,6 @@
 #include "packfile.h"
 #include "submodule-config.h"
 #include "config.h"
-#include "credential.h"
 #include "help.h"
 
 static ssize_t max_tree_entry_len = 4096;
@@ -1048,138 +1047,6 @@ done:
        return ret;
 }
 
-static int starts_with_dot_slash(const char *const path)
-{
-       return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_SLASH |
-                               PATH_MATCH_XPLATFORM);
-}
-
-static int starts_with_dot_dot_slash(const char *const path)
-{
-       return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH |
-                               PATH_MATCH_XPLATFORM);
-}
-
-static int submodule_url_is_relative(const char *url)
-{
-       return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
-}
-
-/*
- * Count directory components that a relative submodule URL should chop
- * from the remote_url it is to be resolved against.
- *
- * In other words, this counts "../" components at the start of a
- * submodule URL.
- *
- * Returns the number of directory components to chop and writes a
- * pointer to the next character of url after all leading "./" and
- * "../" components to out.
- */
-static int count_leading_dotdots(const char *url, const char **out)
-{
-       int result = 0;
-       while (1) {
-               if (starts_with_dot_dot_slash(url)) {
-                       result++;
-                       url += strlen("../");
-                       continue;
-               }
-               if (starts_with_dot_slash(url)) {
-                       url += strlen("./");
-                       continue;
-               }
-               *out = url;
-               return result;
-       }
-}
-/*
- * Check whether a transport is implemented by git-remote-curl.
- *
- * If it is, returns 1 and writes the URL that would be passed to
- * git-remote-curl to the "out" parameter.
- *
- * Otherwise, returns 0 and leaves "out" untouched.
- *
- * Examples:
- *   http::https://example.com/repo.git -> 1, https://example.com/repo.git
- *   https://example.com/repo.git -> 1, https://example.com/repo.git
- *   git://example.com/repo.git -> 0
- *
- * This is for use in checking for previously exploitable bugs that
- * required a submodule URL to be passed to git-remote-curl.
- */
-static int url_to_curl_url(const char *url, const char **out)
-{
-       /*
-        * We don't need to check for case-aliases, "http.exe", and so
-        * on because in the default configuration, is_transport_allowed
-        * prevents URLs with those schemes from being cloned
-        * automatically.
-        */
-       if (skip_prefix(url, "http::", out) ||
-           skip_prefix(url, "https::", out) ||
-           skip_prefix(url, "ftp::", out) ||
-           skip_prefix(url, "ftps::", out))
-               return 1;
-       if (starts_with(url, "http://") ||
-           starts_with(url, "https://") ||
-           starts_with(url, "ftp://") ||
-           starts_with(url, "ftps://")) {
-               *out = url;
-               return 1;
-       }
-       return 0;
-}
-
-static int check_submodule_url(const char *url)
-{
-       const char *curl_url;
-
-       if (looks_like_command_line_option(url))
-               return -1;
-
-       if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
-               char *decoded;
-               const char *next;
-               int has_nl;
-
-               /*
-                * This could be appended to an http URL and url-decoded;
-                * check for malicious characters.
-                */
-               decoded = url_decode(url);
-               has_nl = !!strchr(decoded, '\n');
-
-               free(decoded);
-               if (has_nl)
-                       return -1;
-
-               /*
-                * URLs which escape their root via "../" can overwrite
-                * the host field and previous components, resolving to
-                * URLs like https::example.com/submodule.git and
-                * https:///example.com/submodule.git that were
-                * susceptible to CVE-2020-11008.
-                */
-               if (count_leading_dotdots(url, &next) > 0 &&
-                   (*next == ':' || *next == '/'))
-                       return -1;
-       }
-
-       else if (url_to_curl_url(url, &curl_url)) {
-               struct credential c = CREDENTIAL_INIT;
-               int ret = 0;
-               if (credential_from_url_gently(&c, curl_url, 1) ||
-                   !*c.host)
-                       ret = -1;
-               credential_clear(&c);
-               return ret;
-       }
-
-       return 0;
-}
-
 struct fsck_gitmodules_data {
        const struct object_id *oid;
        struct fsck_options *options;
index 6a48fd12f66f93d132aab6745d67effd482a168f..cbec13b3a20518cb9da1c7d27cde76058d7fe126 100644 (file)
@@ -14,6 +14,8 @@
 #include "parse-options.h"
 #include "thread-utils.h"
 #include "tree-walk.h"
+#include "url.h"
+#include "credential.h"
 
 /*
  * submodule cache lookup structure
@@ -228,6 +230,138 @@ in_component:
        return 0;
 }
 
+static int starts_with_dot_slash(const char *const path)
+{
+       return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_SLASH |
+                               PATH_MATCH_XPLATFORM);
+}
+
+static int starts_with_dot_dot_slash(const char *const path)
+{
+       return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH |
+                               PATH_MATCH_XPLATFORM);
+}
+
+static int submodule_url_is_relative(const char *url)
+{
+       return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
+}
+
+/*
+ * Count directory components that a relative submodule URL should chop
+ * from the remote_url it is to be resolved against.
+ *
+ * In other words, this counts "../" components at the start of a
+ * submodule URL.
+ *
+ * Returns the number of directory components to chop and writes a
+ * pointer to the next character of url after all leading "./" and
+ * "../" components to out.
+ */
+static int count_leading_dotdots(const char *url, const char **out)
+{
+       int result = 0;
+       while (1) {
+               if (starts_with_dot_dot_slash(url)) {
+                       result++;
+                       url += strlen("../");
+                       continue;
+               }
+               if (starts_with_dot_slash(url)) {
+                       url += strlen("./");
+                       continue;
+               }
+               *out = url;
+               return result;
+       }
+}
+/*
+ * Check whether a transport is implemented by git-remote-curl.
+ *
+ * If it is, returns 1 and writes the URL that would be passed to
+ * git-remote-curl to the "out" parameter.
+ *
+ * Otherwise, returns 0 and leaves "out" untouched.
+ *
+ * Examples:
+ *   http::https://example.com/repo.git -> 1, https://example.com/repo.git
+ *   https://example.com/repo.git -> 1, https://example.com/repo.git
+ *   git://example.com/repo.git -> 0
+ *
+ * This is for use in checking for previously exploitable bugs that
+ * required a submodule URL to be passed to git-remote-curl.
+ */
+static int url_to_curl_url(const char *url, const char **out)
+{
+       /*
+        * We don't need to check for case-aliases, "http.exe", and so
+        * on because in the default configuration, is_transport_allowed
+        * prevents URLs with those schemes from being cloned
+        * automatically.
+        */
+       if (skip_prefix(url, "http::", out) ||
+           skip_prefix(url, "https::", out) ||
+           skip_prefix(url, "ftp::", out) ||
+           skip_prefix(url, "ftps::", out))
+               return 1;
+       if (starts_with(url, "http://") ||
+           starts_with(url, "https://") ||
+           starts_with(url, "ftp://") ||
+           starts_with(url, "ftps://")) {
+               *out = url;
+               return 1;
+       }
+       return 0;
+}
+
+int check_submodule_url(const char *url)
+{
+       const char *curl_url;
+
+       if (looks_like_command_line_option(url))
+               return -1;
+
+       if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
+               char *decoded;
+               const char *next;
+               int has_nl;
+
+               /*
+                * This could be appended to an http URL and url-decoded;
+                * check for malicious characters.
+                */
+               decoded = url_decode(url);
+               has_nl = !!strchr(decoded, '\n');
+
+               free(decoded);
+               if (has_nl)
+                       return -1;
+
+               /*
+                * URLs which escape their root via "../" can overwrite
+                * the host field and previous components, resolving to
+                * URLs like https::example.com/submodule.git and
+                * https:///example.com/submodule.git that were
+                * susceptible to CVE-2020-11008.
+                */
+               if (count_leading_dotdots(url, &next) > 0 &&
+                   (*next == ':' || *next == '/'))
+                       return -1;
+       }
+
+       else if (url_to_curl_url(url, &curl_url)) {
+               struct credential c = CREDENTIAL_INIT;
+               int ret = 0;
+               if (credential_from_url_gently(&c, curl_url, 1) ||
+                   !*c.host)
+                       ret = -1;
+               credential_clear(&c);
+               return ret;
+       }
+
+       return 0;
+}
+
 static int name_and_item_from_var(const char *var, struct strbuf *name,
                                  struct strbuf *item)
 {
index 2a37689cc272e30fa6f5dc9cfbbde044a52068c3..b5785686f570820d6b562b10264af25c9f871cb7 100644 (file)
@@ -91,6 +91,9 @@ int config_set_in_gitmodules_file_gently(const char *key, const char *value);
  */
 int check_submodule_name(const char *name);
 
+/* Returns 0 if the URL valid per RFC3986 and -1 otherwise. */
+int check_submodule_url(const char *url);
+
 /*
  * Note: these helper functions exist solely to maintain backward
  * compatibility with 'fetch' and 'update_clone' storing configuration in