Re: [PATCH v2] setup: tighten ownership checks post CVE-2022-24765
- From
- Phillip Wood <phillip.wood123@gmail.com>
- Date
- May 5, 2022, 09:40 UTC
- Message-ID
- <41ec8c09-f31f-46ce-d6ec-4b6fdd78228a@gmail.com>
- In-Reply-To
- <20220505005009.27789-1-carenas@gmail.com>
Hi Carlo
On 05/05/2022 01:50, Carlo Marcelo Arenas Belón wrote:
Show 18 quoted lines
> 8959555cee7 (setup_git_directory(): add an owner check for the top-level > directory, 2022-03-02), adds a function to check for ownership of > repositories using a directory that is representative of it (its workdir) > and ways to add it to an exception list if needed, but that check breaks > when the ownership of the workdir is not the same than the ownership of > directory where the configuration and other relevant files reside. > > An attacker could create a git repository in a directory that he has write > access to but is owned by the victim, and therefore workaround the fix that > was introduced with CVE-2022-24765 to attack them, like in the following > scenario which could result in privilege escalation if root then runs a git > command in that directory or any of its sub directories: > > $ git -C /tmp init > > To avoid that, extend the ensure_valid_ownership function to be able to > check for ownership of both the worktree and the gitdir, and use that for > non bare repositories.
Looking at the code below it now only ever checks the ownership of the gitdir, it no longer checks the ownership of the worktree. I haven't really thought through what happens if I cd into a worktree added by an attacker to a repository that I own which has extentions.worktreeConfig set. My initial thought is that if they can add a worktree then they can probably edit the repository config anyway but I wonder if an attacker can set GIT_COMMON_DIR to a directory where they have write permission to add a worktree to a repository where they don't have write permission.
Show 14 quoted lines
> [...]
> diff --git a/setup.c b/setup.c
> index aad9ace0af9..0fae2d71a3c 100644
> --- a/setup.c
> +++ b/setup.c
> @@ -1054,14 +1054,21 @@ static int safe_directory_cb(const char *key, const char *value, void *d)
> return 0;
> }
>
> -static int ensure_valid_ownership(const char *path)
> +static int ensure_valid_ownership(const char *worktree, const char *gitdir)
> {
> - struct safe_directory_data data = { .path = path };
> + struct safe_directory_data data = { .path = worktree };We keep checking the worktree path against safe.directory - good.
Show 6 quoted lines
> + const char *check_path; > + > + if (gitdir) > + check_path = gitdir; > + else > + check_path = worktree;
We check either the gitdir or worktree but always call this function with a non-NULL gitdir so in fact always check that. This code could be removed.
> if (!git_env_bool("GIT_TEST_ASSUME_DIFFERENT_OWNER", 0) &&
> - is_path_owned_by_current_user(path))
> + is_path_owned_by_current_user(check_path))Previously we checked the ownership of the worktree but now we check the ownership of the gitdir instead to handle the "git -C /tmp init" case.
Show 31 quoted lines
> return 1;
>
> + data.is_safe = 0; /* ensure we are initialized and secure by default */
> read_very_early_config(safe_directory_cb, &data);
>
> return data.is_safe;
> @@ -1166,14 +1173,25 @@ static enum discovery_result setup_git_directory_gently_1(struct strbuf *dir,
> }
> strbuf_setlen(dir, offset);
> if (gitdirenv) {
> - if (!ensure_valid_ownership(dir->buf))
> + const char *gitdir_to_check = gitdirenv;
> + struct strbuf gdbuf = STRBUF_INIT;
> + int ret;
> +
> + if (!is_absolute_path(gitdirenv)) {
> + strbuf_addf(&gdbuf, "%s/%s", dir->buf,
> + gitdirenv);
> + gitdir_to_check = gdbuf.buf;
> + }
> + ret = ensure_valid_ownership(dir->buf, gitdir_to_check);
> + strbuf_release(&gdbuf);
> + if (!ret)
> return GIT_DIR_INVALID_OWNERSHIP;
> strbuf_addstr(gitdir, gitdirenv);
> return GIT_DIR_DISCOVERED;
> }
>
> if (is_git_directory(dir->buf)) {
> - if (!ensure_valid_ownership(dir->buf))
> + if (!ensure_valid_ownership(NULL, dir->buf))Previously we checked bare repositories against safe.directory now we no longer do that as worktree is NULL. If this is intentional it should be flagged up in the commit message.
Best Wishes
Phillip
Show 43 quoted lines
> return GIT_DIR_INVALID_OWNERSHIP; > strbuf_addstr(gitdir, "."); > return GIT_DIR_BARE; > diff --git a/t/t0034-root-safe-directory.sh b/t/t0034-root-safe-directory.sh > index a68e1d7602b..a3ddebb009a 100755 > --- a/t/t0034-root-safe-directory.sh > +++ b/t/t0034-root-safe-directory.sh > @@ -47,6 +47,35 @@ test_expect_success SUDO 'sudo git status as original owner' ' > ) > ' > > +test_expect_success SUDO 'unsecure worktree with non bare repository' ' > + sudo rm -rf root && > + sudo mkdir -p root/t && > + sudo chmod 1777 root/t && > + ( > + cd root/t && > + git init && > + git status && > + sudo git status && > + run_with_sudo <<-END > + unset SUDO_UID && > + ! git status > + END > + ) > +' > + > +test_expect_success SUDO 'non bare repository using a gitfile' ' > + sudo rm -rf root && > + mkdir -p root/w && > + mkdir -p root/e && > + ( > + cd root/w && > + git init --separate-git-dir ../e && > + git status && > + sudo chown -R root ../e && > + test_must_fail git status > + ) > +' > + > # this destroys the test environment used above > test_expect_success SUDO 'cleanup regression' ' > sudo rm -rf root