git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: "Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
Cc: git@vger.kernel.org, Jeff King <peff@peff.net>
Subject: Re: [PATCH v2] upload-pack: allow shallow fetching from a read-only repository
Date: Tue, 04 Mar 2014 10:10:22 -0800	[thread overview]
Message-ID: <xmqqr46hltrl.fsf@gitster.dls.corp.google.com> (raw)
In-Reply-To: <1393936205-15953-1-git-send-email-pclouds@gmail.com> ("Nguyễn	Thái Ngọc Duy"'s message of "Tue, 4 Mar 2014 19:30:05 +0700")

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Before cdab485 (upload-pack: delegate rev walking in shallow fetch to
> pack-objects - 2013-08-16) upload-pack does not write to the source
> repository. cdab485 starts to write $GIT_DIR/shallow_XXXXXX if it's a
> shallow fetch, so the source repo must be writable.
>
> git:// servers do not need write access to repos and usually don't,
> which mean cdab485 breaks shallow clone over git://
>
> Fall back to $TMPDIR if $GIT_DIR/shallow_XXXXXX cannot be created in
> this case. Note that in other cases that write $GIT_DIR/shallow_XXXXXX
> and eventually rename it to $GIT_DIR/shallow, there is no fallback to
> $TMPDIR.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  Rebased on top of jk/shallow-update-fix

Hmph.

I notice that the original code, with or without this change, allows
upload-pack spawned by daemon to attempt to write into GIT_DIR.
As upload-pack is supposed to be a read-only operation, this is
quite bad.

Perhaps we should give server operators an option to run their
daemon -> upload-pack chain to always write to a throw-away
directory of their choice, without ever attempting to write to
GIT_DIR it serves?

How well is the access to the temporary shallow file controlled in
your code (sorry, but I do not recall carefully reading your patch
that added the mechanism with security issues in mind, so now I am
asking)?  When it is redirected to TMPDIR (let's forget GIT_DIR for
now---if an attacker can write into there, the repository is already
lost), can an attacker race with us to cause us to overwrite we do
not expect to?

Even if it turns out that this patch is secure enough as-is, we
definitely need to make sure that server operators, who want to keep
their upload-pack truly a read-only operation, know that it is
necessary to (1) keep the system user they run git-daemon under
incapable of writing into GIT_DIR, and (2) make sure TMPDIR points
at somewhere only git-daemon user and nobody else can write into,
somewhere in the documentation.

> diff --git a/fetch-pack.c b/fetch-pack.c
> index ae8550e..b71d186 100644
> --- a/fetch-pack.c
> +++ b/fetch-pack.c
> @@ -853,7 +853,7 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
>  		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
>  					NULL);
>  	else if (si->nr_ours || si->nr_theirs)
> -		alternate_shallow_file = setup_temporary_shallow(si->shallow);
> +		alternate_shallow_file = setup_temporary_shallow(si->shallow, 0);
>  	else
>  		alternate_shallow_file = NULL;
>  	if (get_pack(args, fd, pack_lockfile))
> diff --git a/shallow.c b/shallow.c
> index c7602ce..ad28af6 100644
> --- a/shallow.c
> +++ b/shallow.c
> @@ -224,7 +224,8 @@ static void remove_temporary_shallow_on_signal(int signo)
>  	raise(signo);
>  }
>  
> -const char *setup_temporary_shallow(const struct sha1_array *extra)
> +const char *setup_temporary_shallow(const struct sha1_array *extra,
> +				    int read_only)
>  {
>  	static int installed_handler;
>  	struct strbuf sb = STRBUF_INIT;
> @@ -235,7 +236,15 @@ const char *setup_temporary_shallow(const struct sha1_array *extra)
>  
>  	if (write_shallow_commits(&sb, 0, extra)) {
>  		strbuf_addstr(&temporary_shallow, git_path("shallow_XXXXXX"));
> -		fd = xmkstemp(temporary_shallow.buf);
> +		fd = mkstemp(temporary_shallow.buf);
> +		if (read_only && fd < 0) {
> +			strbuf_grow(&temporary_shallow, PATH_MAX);
> +			fd = git_mkstemp(temporary_shallow.buf, PATH_MAX,
> +					 "shallow_XXXXXX");
> +		}
> +		if (fd < 0)
> +			die_errno("Unable to create temporary file '%s'",
> +				  temporary_shallow.buf);
>  
>  		if (!installed_handler) {
>  			atexit(remove_temporary_shallow);
> diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
> index b0fa738..171db88 100755
> --- a/t/t5537-fetch-shallow.sh
> +++ b/t/t5537-fetch-shallow.sh
> @@ -173,6 +173,19 @@ EOF
>  	)
>  '
>  
> +test_expect_success POSIXPERM 'shallow fetch from a read-only repo' '

s/POSIXPERM/&,SANITY/, perhaps?

Thinking of it again, perhaps POSIXPERM should imply SANITY is required?

> +	cp -R .git read-only.git &&
> +	find read-only.git -print | xargs chmod -w &&
> +	test_when_finished "find read-only.git -type d -print | xargs chmod +w" &&
> +	git clone --no-local --depth=2 read-only.git from-read-only &&
> +	git --git-dir=from-read-only/.git log --format=%s >actual &&
> +	cat >expect <<EOF &&
> +add-1-back
> +4
> +EOF
> +	test_cmp expect actual
> +'
> +
>  if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
>  	say 'skipping remaining tests, git built without http support'
>  	test_done
> diff --git a/upload-pack.c b/upload-pack.c
> index a3c52f6..b538f32 100644
> --- a/upload-pack.c
> +++ b/upload-pack.c
> @@ -84,7 +84,7 @@ static void create_pack_file(void)
>  	const char *shallow_file = NULL;
>  
>  	if (shallow_nr) {
> -		shallow_file = setup_temporary_shallow(NULL);
> +		shallow_file = setup_temporary_shallow(NULL, 1);
>  		argv[arg++] = "--shallow-file";
>  		argv[arg++] = shallow_file;
>  	}

  reply	other threads:[~2014-03-04 18:10 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-02-27  7:13 [PATCH] upload-pack: allow shallow fetching from a read-only repository Nguyễn Thái Ngọc Duy
2014-02-27  9:04 ` Jeff King
2014-02-27  9:10   ` [PATCH] shallow: verify shallow file after taking lock Jeff King
2014-02-27  9:22     ` Jeff King
2014-02-27 10:18       ` Duy Nguyen
2014-02-27 10:56         ` [PATCH] shallow: use stat_validity to check for up-to-date file Jeff King
2014-02-27 10:11   ` [PATCH] upload-pack: allow shallow fetching from a read-only repository Duy Nguyen
2014-02-27 11:25     ` [PATCH] shallow: automatically clean up shallow tempfiles Jeff King
2014-03-04 12:30 ` [PATCH v2] upload-pack: allow shallow fetching from a read-only repository Nguyễn Thái Ngọc Duy
2014-03-04 18:10   ` Junio C Hamano [this message]
2014-03-05 12:43     ` Duy Nguyen
2014-03-05 19:50       ` Junio C Hamano
2014-03-06  8:49   ` [PATCH v3] upload-pack: send shallow info over stdin to pack-objects Nguyễn Thái Ngọc Duy
2014-03-06 18:37     ` Junio C Hamano
2014-03-06 23:13       ` Duy Nguyen
2014-03-07 18:27         ` Junio C Hamano
2014-03-08  0:08           ` Duy Nguyen
2014-03-10 15:23             ` Junio C Hamano
2014-03-07  1:24     ` Duy Nguyen
2014-03-07 18:33       ` Junio C Hamano
2014-03-11 12:59     ` [PATCH v4] " Nguyễn Thái Ngọc Duy

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=xmqqr46hltrl.fsf@gitster.dls.corp.google.com \
    --to=gitster@pobox.com \
    --cc=git@vger.kernel.org \
    --cc=pclouds@gmail.com \
    --cc=peff@peff.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).