transport: add protocol policy config option

Previously the `GIT_ALLOW_PROTOCOL` environment variable was used to specify a whitelist of protocols to be used in clone/fetch/push commands. This patch introduces new configuration options for more fine-grained control for allowing/disallowing protocols. This also has the added benefit of allowing easier construction of a protocol whitelist on systems where setting an environment variable is non-trivial. Now users can specify a policy to be used for each type of protocol via the 'protocol.<name>.allow' config option. A default policy for all unconfigured protocols can be set with the 'protocol.allow' config option. If no user configured default is made git will allow known-safe protocols (http, https, git, ssh, file), disallow known-dangerous protocols (ext), and have a default policy of `user` for all other protocols. The supported policies are `always`, `never`, and `user`. The `user` policy can be used to configure a protocol to be usable when explicitly used by a user, while disallowing it for commands which run clone/fetch/push commands without direct user intervention (e.g. recursive initialization of submodules). Commands which can potentially clone/fetch/push from untrusted repositories without user intervention can export `GIT_PROTOCOL_FROM_USER` with a value of '0' to prevent protocols configured to the `user` policy from being used. Fix remote-ext tests to use the new config to allow the ext protocol to be tested. Based on a patch by Jeff King <peff@peff.net> Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Brandon Williams committed Dec 14, 2016 at 14:39 UTC f1762d772e9b415a3163abf5f217fc3b71a3b40e
7 files changed +264 -39
Documentation/config.txt
+46
@@ -2260,6 +2260,52 @@ pretty.<name>::
2260 Note that an alias with the same name as a built-in format
2261 will be silently ignored.
2262
2263 +protocol.allow::
2264 + If set, provide a user defined default policy for all protocols which
2265 + don't explicitly have a policy (`protocol.<name>.allow`). By default,
2266 + if unset, known-safe protocols (http, https, git, ssh, file) have a
2267 + default policy of `always`, known-dangerous protocols (ext) have a
2268 + default policy of `never`, and all other protocols have a default
2269 + policy of `user`. Supported policies:
2270 ++
2271 +--
2272 +
2273 +* `always` - protocol is always able to be used.
2274 +
2275 +* `never` - protocol is never able to be used.
2276 +
2277 +* `user` - protocol is only able to be used when `GIT_PROTOCOL_FROM_USER` is
2278 + either unset or has a value of 1. This policy should be used when you want a
2279 + protocol to be directly usable by the user but don't want it used by commands which
2280 + execute clone/fetch/push commands without user input, e.g. recursive
2281 + submodule initialization.
2282 +
2283 +--
2284 +
2285 +protocol.<name>.allow::
2286 + Set a policy to be used by protocol `<name>` with clone/fetch/push
2287 + commands. See `protocol.allow` above for the available policies.
2288 ++
2289 +The protocol names currently used by git are:
2290 ++
2291 +--
2292 + - `file`: any local file-based path (including `file://` URLs,
2293 + or local paths)
2294 +
2295 + - `git`: the anonymous git protocol over a direct TCP
2296 + connection (or proxy, if configured)
2297 +
2298 + - `ssh`: git over ssh (including `host:path` syntax,
2299 + `ssh://`, etc).
2300 +
2301 + - `http`: git over http, both "smart http" and "dumb http".
2302 + Note that this does _not_ include `https`; if you want to configure
2303 + both, you must do so individually.
2304 +
2305 + - any external helpers are named by their protocol (e.g., use
2306 + `hg` to allow the `git-remote-hg` helper)
2307 +--
2308 +
2309 pull.ff::
2310 By default, Git does not create an extra merge commit when merging
2311 a commit that is a descendant of the current commit. Instead, the
Documentation/git.txt
+14 -24
@@ -1129,30 +1129,20 @@ of clones and fetches.
1129 cloning a repository to make a backup).
1130
1131 `GIT_ALLOW_PROTOCOL`::
1132 - If set, provide a colon-separated list of protocols which are
1133 - allowed to be used with fetch/push/clone. This is useful to
1134 - restrict recursive submodule initialization from an untrusted
1135 - repository. Any protocol not mentioned will be disallowed (i.e.,
1136 - this is a whitelist, not a blacklist). If the variable is not
1137 - set at all, all protocols are enabled. The protocol names
1138 - currently used by git are:
1139 -
1140 - - `file`: any local file-based path (including `file://` URLs,
1141 - or local paths)
1142 -
1143 - - `git`: the anonymous git protocol over a direct TCP
1144 - connection (or proxy, if configured)
1145 -
1146 - - `ssh`: git over ssh (including `host:path` syntax,
1147 - `ssh://`, etc).
1148 -
1149 - - `http`: git over http, both "smart http" and "dumb http".
1150 - Note that this does _not_ include `https`; if you want both,
1151 - you should specify both as `http:https`.
1152 -
1153 - - any external helpers are named by their protocol (e.g., use
1154 - `hg` to allow the `git-remote-hg` helper)
1155 -
1132 + If set to a colon-separated list of protocols, behave as if
1133 + `protocol.allow` is set to `never`, and each of the listed
1134 + protocols has `protocol.<name>.allow` set to `always`
1135 + (overriding any existing configuration). In other words, any
1136 + protocol not mentioned will be disallowed (i.e., this is a
1137 + whitelist, not a blacklist). See the description of
1138 + `protocol.allow` in linkgit:git-config[1] for more details.
1139 +
1140 +`GIT_PROTOCOL_FROM_USER`::
1141 + Set to 0 to prevent protocols used by fetch/push/clone which are
1142 + configured to the `user` state. This is useful to restrict recursive
1143 + submodule initialization from an untrusted repository or for programs
1144 + which feed potentially-untrusted URLS to git commands. See
1145 + linkgit:git-config[1] for more details.
1146
1147 Discussion[[Discussion]]
1148 ------------------------
git-submodule.sh
+4 -8
@@ -22,14 +22,10 @@ require_work_tree
22 wt_prefix=$(git rev-parse --show-prefix)
23 cd_to_toplevel
24
25 -# Restrict ourselves to a vanilla subset of protocols; the URLs
26 -# we get are under control of a remote repository, and we do not
27 -# want them kicking off arbitrary git-remote-* programs.
28 -#
29 -# If the user has already specified a set of allowed protocols,
30 -# we assume they know what they're doing and use that instead.
31 -: ${GIT_ALLOW_PROTOCOL=file:git:http:https:ssh}
32 -export GIT_ALLOW_PROTOCOL
25 +# Tell the rest of git that any URLs we get don't come
26 +# directly from the user, so it can apply policy as appropriate.
27 +GIT_PROTOCOL_FROM_USER=0
28 +export GIT_PROTOCOL_FROM_USER
29
30 command=
31 branch=
t/lib-proto-disable.sh
+125 -5
@@ -1,10 +1,7 @@
1 # Test routines for checking protocol disabling.
2
3 -# test cloning a particular protocol
4 -# $1 - description of the protocol
5 -# $2 - machine-readable name of the protocol
6 -# $3 - the URL to try cloning
7 -test_proto () {
3 +# Test clone/fetch/push with GIT_ALLOW_PROTOCOL whitelist
4 +test_whitelist () {
5 desc=$1
6 proto=$2
7 url=$3
@@ -62,6 +59,129 @@ test_proto () {
59 test_must_fail git clone --bare "$url" tmp.git
60 )
61 '
62 +
63 + test_expect_success "clone $desc (env var has precedence)" '
64 + rm -rf tmp.git &&
65 + (
66 + GIT_ALLOW_PROTOCOL=none &&
67 + export GIT_ALLOW_PROTOCOL &&
68 + test_must_fail git -c protocol.allow=always clone --bare "$url" tmp.git &&
69 + test_must_fail git -c protocol.$proto.allow=always clone --bare "$url" tmp.git
70 + )
71 + '
72 +}
73 +
74 +test_config () {
75 + desc=$1
76 + proto=$2
77 + url=$3
78 +
79 + # Test clone/fetch/push with protocol.<type>.allow config
80 + test_expect_success "clone $desc (enabled with config)" '
81 + rm -rf tmp.git &&
82 + git -c protocol.$proto.allow=always clone --bare "$url" tmp.git
83 + '
84 +
85 + test_expect_success "fetch $desc (enabled)" '
86 + git -C tmp.git -c protocol.$proto.allow=always fetch
87 + '
88 +
89 + test_expect_success "push $desc (enabled)" '
90 + git -C tmp.git -c protocol.$proto.allow=always push origin HEAD:pushed
91 + '
92 +
93 + test_expect_success "push $desc (disabled)" '
94 + test_must_fail git -C tmp.git -c protocol.$proto.allow=never push origin HEAD:pushed
95 + '
96 +
97 + test_expect_success "fetch $desc (disabled)" '
98 + test_must_fail git -C tmp.git -c protocol.$proto.allow=never fetch
99 + '
100 +
101 + test_expect_success "clone $desc (disabled)" '
102 + rm -rf tmp.git &&
103 + test_must_fail git -c protocol.$proto.allow=never clone --bare "$url" tmp.git
104 + '
105 +
106 + # Test clone/fetch/push with protocol.user.allow and its env var
107 + test_expect_success "clone $desc (enabled)" '
108 + rm -rf tmp.git &&
109 + git -c protocol.$proto.allow=user clone --bare "$url" tmp.git
110 + '
111 +
112 + test_expect_success "fetch $desc (enabled)" '
113 + git -C tmp.git -c protocol.$proto.allow=user fetch
114 + '
115 +
116 + test_expect_success "push $desc (enabled)" '
117 + git -C tmp.git -c protocol.$proto.allow=user push origin HEAD:pushed
118 + '
119 +
120 + test_expect_success "push $desc (disabled)" '
121 + (
122 + cd tmp.git &&
123 + GIT_PROTOCOL_FROM_USER=0 &&
124 + export GIT_PROTOCOL_FROM_USER &&
125 + test_must_fail git -c protocol.$proto.allow=user push origin HEAD:pushed
126 + )
127 + '
128 +
129 + test_expect_success "fetch $desc (disabled)" '
130 + (
131 + cd tmp.git &&
132 + GIT_PROTOCOL_FROM_USER=0 &&
133 + export GIT_PROTOCOL_FROM_USER &&
134 + test_must_fail git -c protocol.$proto.allow=user fetch
135 + )
136 + '
137 +
138 + test_expect_success "clone $desc (disabled)" '
139 + rm -rf tmp.git &&
140 + (
141 + GIT_PROTOCOL_FROM_USER=0 &&
142 + export GIT_PROTOCOL_FROM_USER &&
143 + test_must_fail git -c protocol.$proto.allow=user clone --bare "$url" tmp.git
144 + )
145 + '
146 +
147 + # Test clone/fetch/push with protocol.allow user defined default
148 + test_expect_success "clone $desc (enabled)" '
149 + rm -rf tmp.git &&
150 + git config --global protocol.allow always &&
151 + git clone --bare "$url" tmp.git
152 + '
153 +
154 + test_expect_success "fetch $desc (enabled)" '
155 + git -C tmp.git fetch
156 + '
157 +
158 + test_expect_success "push $desc (enabled)" '
159 + git -C tmp.git push origin HEAD:pushed
160 + '
161 +
162 + test_expect_success "push $desc (disabled)" '
163 + git config --global protocol.allow never &&
164 + test_must_fail git -C tmp.git push origin HEAD:pushed
165 + '
166 +
167 + test_expect_success "fetch $desc (disabled)" '
168 + test_must_fail git -C tmp.git fetch
169 + '
170 +
171 + test_expect_success "clone $desc (disabled)" '
172 + rm -rf tmp.git &&
173 + test_must_fail git clone --bare "$url" tmp.git
174 + '
175 +}
176 +
177 +# test cloning a particular protocol
178 +# $1 - description of the protocol
179 +# $2 - machine-readable name of the protocol
180 +# $3 - the URL to try cloning
181 +test_proto () {
182 + test_whitelist "$@"
183 +
184 + test_config "$@"
185 }
186
187 # set up an ssh wrapper that will access $host/$repo in the
t/t5509-fetch-push-namespaces.sh
+1
@@ -4,6 +4,7 @@ test_description='fetch/push involving ref namespaces'
4 . ./test-lib.sh
5
6 test_expect_success setup '
7 + git config --global protocol.ext.allow user &&
8 test_tick &&
9 git init original &&
10 (
t/t5802-connect-helper.sh
+1
@@ -4,6 +4,7 @@ test_description='ext::cmd remote "connect" helper'
4 . ./test-lib.sh
5
6 test_expect_success setup '
7 + git config --global protocol.ext.allow user &&
8 test_tick &&
9 git commit --allow-empty -m initial &&
10 test_tick &&
transport.c
+73 -2
@@ -617,10 +617,81 @@ static const struct string_list *protocol_whitelist(void)
617 return enabled ? &allowed : NULL;
618 }
619
620 +enum protocol_allow_config {
621 + PROTOCOL_ALLOW_NEVER = 0,
622 + PROTOCOL_ALLOW_USER_ONLY,
623 + PROTOCOL_ALLOW_ALWAYS
624 +};
625 +
626 +static enum protocol_allow_config parse_protocol_config(const char *key,
627 + const char *value)
628 +{
629 + if (!strcasecmp(value, "always"))
630 + return PROTOCOL_ALLOW_ALWAYS;
631 + else if (!strcasecmp(value, "never"))
632 + return PROTOCOL_ALLOW_NEVER;
633 + else if (!strcasecmp(value, "user"))
634 + return PROTOCOL_ALLOW_USER_ONLY;
635 +
636 + die("unknown value for config '%s': %s", key, value);
637 +}
638 +
639 +static enum protocol_allow_config get_protocol_config(const char *type)
640 +{
641 + char *key = xstrfmt("protocol.%s.allow", type);
642 + char *value;
643 +
644 + /* first check the per-protocol config */
645 + if (!git_config_get_string(key, &value)) {
646 + enum protocol_allow_config ret =
647 + parse_protocol_config(key, value);
648 + free(key);
649 + free(value);
650 + return ret;
651 + }
652 + free(key);
653 +
654 + /* if defined, fallback to user-defined default for unknown protocols */
655 + if (!git_config_get_string("protocol.allow", &value)) {
656 + enum protocol_allow_config ret =
657 + parse_protocol_config("protocol.allow", value);
658 + free(value);
659 + return ret;
660 + }
661 +
662 + /* fallback to built-in defaults */
663 + /* known safe */
664 + if (!strcmp(type, "http") ||
665 + !strcmp(type, "https") ||
666 + !strcmp(type, "git") ||
667 + !strcmp(type, "ssh") ||
668 + !strcmp(type, "file"))
669 + return PROTOCOL_ALLOW_ALWAYS;
670 +
671 + /* known scary; err on the side of caution */
672 + if (!strcmp(type, "ext"))
673 + return PROTOCOL_ALLOW_NEVER;
674 +
675 + /* unknown; by default let them be used only directly by the user */
676 + return PROTOCOL_ALLOW_USER_ONLY;
677 +}
678 +
679 int is_transport_allowed(const char *type)
680 {
622 - const struct string_list *allowed = protocol_whitelist();
623 - return !allowed || string_list_has_string(allowed, type);
681 + const struct string_list *whitelist = protocol_whitelist();
682 + if (whitelist)
683 + return string_list_has_string(whitelist, type);
684 +
685 + switch (get_protocol_config(type)) {
686 + case PROTOCOL_ALLOW_ALWAYS:
687 + return 1;
688 + case PROTOCOL_ALLOW_NEVER:
689 + return 0;
690 + case PROTOCOL_ALLOW_USER_ONLY:
691 + return git_env_bool("GIT_PROTOCOL_FROM_USER", 1);
692 + }
693 +
694 + die("BUG: invalid protocol_allow_config type");
695 }
696
697 void transport_check_allowed(const char *type)