In v2.20.x, Git clones submodules recursively by first creating the
submodules' gitdirs and _then_ "updating" the submodules. This can lead
to the situation where the clone path is taken because the directory
(while it exists already) is not a git directory, but then the clone
fails because that gitdir is unexpectedly already a directory.
This _also_ works around the vulnerability that was fixed in "Disallow
dubiously-nested submodule git directories", but it produces a different
error message than the one expected by the test case, therefore we
adjust the test case accordingly.
Note: as the two submodules "race each other", there are actually two
possible error messages, therefore we have to teach the test case to
expect _two_ possible (and good) outcomes in addition to the one it
expected before.
Note: this workaround is only necessary for the v2.20.x release train;
The behavior changed again in v2.21.x so that the original test case's
expectations are met again.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
* maint-2.18: (33 commits)
Git 2.18.2
Git 2.17.3
Git 2.16.6
test-drop-caches: use `has_dos_drive_prefix()`
Git 2.15.4
Git 2.14.6
mingw: handle `subst`-ed "DOS drives"
mingw: refuse to access paths with trailing spaces or periods
mingw: refuse to access paths with illegal characters
unpack-trees: let merged_entry() pass through do_add_entry()'s errors
quote-stress-test: offer to test quoting arguments for MSYS2 sh
t6130/t9350: prepare for stringent Win32 path validation
quote-stress-test: allow skipping some trials
quote-stress-test: accept arguments to test via the command-line
tests: add a helper to stress test argument quoting
mingw: fix quoting of arguments
Disallow dubiously-nested submodule git directories
protect_ntfs: turn on NTFS protection by default
path: also guard `.gitmodules` against NTFS Alternate Data Streams
is_ntfs_dotgit(): speed it up
...
* maint-2.17: (32 commits)
Git 2.17.3
Git 2.16.6
test-drop-caches: use `has_dos_drive_prefix()`
Git 2.15.4
Git 2.14.6
mingw: handle `subst`-ed "DOS drives"
mingw: refuse to access paths with trailing spaces or periods
mingw: refuse to access paths with illegal characters
unpack-trees: let merged_entry() pass through do_add_entry()'s errors
quote-stress-test: offer to test quoting arguments for MSYS2 sh
t6130/t9350: prepare for stringent Win32 path validation
quote-stress-test: allow skipping some trials
quote-stress-test: accept arguments to test via the command-line
tests: add a helper to stress test argument quoting
mingw: fix quoting of arguments
Disallow dubiously-nested submodule git directories
protect_ntfs: turn on NTFS protection by default
path: also guard `.gitmodules` against NTFS Alternate Data Streams
is_ntfs_dotgit(): speed it up
mingw: disallow backslash characters in tree objects' file names
...
* maint-2.16: (31 commits)
Git 2.16.6
test-drop-caches: use `has_dos_drive_prefix()`
Git 2.15.4
Git 2.14.6
mingw: handle `subst`-ed "DOS drives"
mingw: refuse to access paths with trailing spaces or periods
mingw: refuse to access paths with illegal characters
unpack-trees: let merged_entry() pass through do_add_entry()'s errors
quote-stress-test: offer to test quoting arguments for MSYS2 sh
t6130/t9350: prepare for stringent Win32 path validation
quote-stress-test: allow skipping some trials
quote-stress-test: accept arguments to test via the command-line
tests: add a helper to stress test argument quoting
mingw: fix quoting of arguments
Disallow dubiously-nested submodule git directories
protect_ntfs: turn on NTFS protection by default
path: also guard `.gitmodules` against NTFS Alternate Data Streams
is_ntfs_dotgit(): speed it up
mingw: disallow backslash characters in tree objects' file names
path: safeguard `.git` against NTFS Alternate Streams Accesses
...
On Windows, filenames cannot have trailing spaces or periods, when
opening such paths, they are stripped automatically. Read: you can open
the file `README` via the file name `README . . .`. This ambiguity can
be used in combination with other security bugs to cause e.g. remote
code execution during recursive clones. This patch series fixes that.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When creating a directory on Windows whose path ends in a space or a
period (or chains thereof), the Win32 API "helpfully" trims those. For
example, `mkdir("abc ");` will return success, but actually create a
directory called `abc` instead.
This stems back to the DOS days, when all file names had exactly 8
characters plus exactly 3 characters for the file extension, and the
only way to have shorter names was by padding with spaces.
Sadly, this "helpful" behavior is a bit inconsistent: after a successful
`mkdir("abc ");`, a `mkdir("abc /def")` will actually _fail_ (because
the directory `abc ` does not actually exist).
Even if it would work, we now have a serious problem because a Git
repository could contain directories `abc` and `abc `, and on Windows,
they would be "merged" unintentionally.
As these paths are illegal on Windows, anyway, let's disallow any
accesses to such paths on that Operating System.
For practical reasons, this behavior is still guarded by the
config setting `core.protectNTFS`: it is possible (and at least two
regression tests make use of it) to create commits without involving the
worktree. In such a scenario, it is of course possible -- even on
Windows -- to create such file names.
Among other consequences, this patch disallows submodules' paths to end
in spaces on Windows (which would formerly have confused Git enough to
try to write into incorrect paths, anyway).
While this patch does not fix a vulnerability on its own, it prevents an
attack vector that was exploited in demonstrations of a number of
recently-fixed security bugs.
The regression test added to `t/t7417-submodule-path-url.sh` reflects
that attack vector.
Note that we have to adjust the test case "prevent git~1 squatting on
Windows" in `t/t7415-submodule-names.sh` because of a very subtle issue.
It tries to clone two submodules whose names differ only in a trailing
period character, and as a consequence their git directories differ in
the same way. Previously, when Git tried to clone the second submodule,
it thought that the git directory already existed (because on Windows,
when you create a directory with the name `b.` it actually creates `b`),
but with this patch, the first submodule's clone will fail because of
the illegal name of the git directory. Therefore, when cloning the
second submodule, Git will take a different code path: a fresh clone
(without an existing git directory). Both code paths fail to clone the
second submodule, both because the the corresponding worktree directory
exists and is not empty, but the error messages are worded differently.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Currently it is technically possible to let a submodule's git
directory point right into the git dir of a sibling submodule.
Example: the git directories of two submodules with the names `hippo`
and `hippo/hooks` would be `.git/modules/hippo/` and
`.git/modules/hippo/hooks/`, respectively, but the latter is already
intended to house the former's hooks.
In most cases, this is just confusing, but there is also a (quite
contrived) attack vector where Git can be fooled into mistaking remote
content for file contents it wrote itself during a recursive clone.
Let's plug this bug.
To do so, we introduce the new function `validate_submodule_git_dir()`
which simply verifies that no git dir exists for any leading directories
of the submodule name (if there are any).
Note: this patch specifically continues to allow sibling modules names
of the form `core/lib`, `core/doc`, etc, as long as `core` is not a
submodule name.
This fixes CVE-2019-1387.
Reported-by: Nicolas Joly <Nicolas.Joly@microsoft.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The backslash character is not a valid part of a file name on Windows.
Hence it is dangerous to allow writing files that were unpacked from
tree objects, when the stored file name contains a backslash character:
it will be misinterpreted as directory separator.
This not only causes ambiguity when a tree contains a blob `a\b` and a
tree `a` that contains a blob `b`, but it also can be used as part of an
attack vector to side-step the careful protections against writing into
the `.git/` directory during a clone of a maliciously-crafted
repository.
Let's prevent that, addressing CVE-2019-1354.
Note: we guard against backslash characters in tree objects' file names
_only_ on Windows (because on other platforms, even on those where NTFS
volumes can be mounted, the backslash character is _not_ a directory
separator), and _only_ when `core.protectNTFS = true` (because users
might need to generate tree objects for other platforms, of course
without touching the worktree, e.g. using `git update-index
--cacheinfo`).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
In addition to preventing `.git` from being tracked by Git, on Windows
we also have to prevent `git~1` from being tracked, as the default NTFS
short name (also known as the "8.3 filename") for the file name `.git`
is `git~1`, otherwise it would be possible for malicious repositories to
write directly into the `.git/` directory, e.g. a `post-checkout` hook
that would then be executed _during_ a recursive clone.
When we implemented appropriate protections in 2b4c6efc82 (read-cache:
optionally disallow NTFS .git variants, 2014-12-16), we had analyzed
carefully that the `.git` directory or file would be guaranteed to be
the first directory entry to be written. Otherwise it would be possible
e.g. for a file named `..git` to be assigned the short name `git~1` and
subsequently, the short name generated for `.git` would be `git~2`. Or
`git~3`. Or even `~9999999` (for a detailed explanation of the lengths
we have to go to protect `.gitmodules`, see the commit message of
e7cb0b4455 (is_ntfs_dotgit: match other .git files, 2018-05-11)).
However, by exploiting two issues (that will be addressed in a related
patch series close by), it is currently possible to clone a submodule
into a non-empty directory:
- On Windows, file names cannot end in a space or a period (for
historical reasons: the period separating the base name from the file
extension was not actually written to disk, and the base name/file
extension was space-padded to the full 8/3 characters, respectively).
Helpfully, when creating a directory under the name, say, `sub.`, that
trailing period is trimmed automatically and the actual name on disk
is `sub`.
This means that while Git thinks that the submodule names `sub` and
`sub.` are different, they both access `.git/modules/sub/`.
- While the backslash character is a valid file name character on Linux,
it is not so on Windows. As Git tries to be cross-platform, it
therefore allows backslash characters in the file names stored in tree
objects.
Which means that it is totally possible that a submodule `c` sits next
to a file `c\..git`, and on Windows, during recursive clone a file
called `..git` will be written into `c/`, of course _before_ the
submodule is cloned.
Note that the actual exploit is not quite as simple as having a
submodule `c` next to a file `c\..git`, as we have to make sure that the
directory `.git/modules/b` already exists when the submodule is checked
out, otherwise a different code path is taken in `module_clone()` that
does _not_ allow a non-empty submodule directory to exist already.
Even if we will address both issues nearby (the next commit will
disallow backslash characters in tree entries' file names on Windows,
and another patch will disallow creating directories/files with trailing
spaces or periods), it is a wise idea to defend in depth against this
sort of attack vector: when submodules are cloned recursively, we now
_require_ the directory to be empty, addressing CVE-2019-1349.
Note: the code path we patch is shared with the code path of `git
submodule update --init`, which must not expect, in general, that the
directory is empty. Hence we have to introduce the new option
`--force-init` and hand it all the way down from `git submodule` to the
actual `git submodule--helper` process that performs the initial clone.
Reported-by: Nicolas Joly <Nicolas.Joly@microsoft.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
We added an fsck check in ed8b10f631 (fsck: check
.gitmodules content, 2018-05-02) as a defense against the
vulnerability from 0383bbb901 (submodule-config: verify
submodule names as paths, 2018-04-30). With the idea that
up-to-date hosting sites could protect downstream unpatched
clients that fetch from them.
As part of that defense, we reject any ".gitmodules" entry
that is not syntactically valid. The theory is that if we
cannot even parse the file, we cannot accurately check it
for vulnerabilities. And anybody with a broken .gitmodules
file would eventually want to know anyway.
But there are a few reasons this is a bad tradeoff in
practice:
- for this particular vulnerability, the client has to be
able to parse the file. So you cannot sneak an attack
through using a broken file, assuming the config parsers
for the process running fsck and the eventual victim are
functionally equivalent.
- a broken .gitmodules file is not necessarily a problem.
Our fsck check detects .gitmodules in _any_ tree, not
just at the root. And the presence of a .gitmodules file
does not necessarily mean it will be used; you'd have to
also have gitlinks in the tree. The cgit repository, for
example, has a file named .gitmodules from a
pre-submodule attempt at sharing code, but does not
actually have any gitlinks.
- when the fsck check is used to reject a push, it's often
hard to work around. The pusher may not have full control
over the destination repository (e.g., if it's on a
hosting server, they may need to contact the hosting
site's support). And the broken .gitmodules may be too
far back in history for rewriting to be feasible (again,
this is an issue for cgit).
So we're being unnecessarily restrictive without actually
improving the security in a meaningful way. It would be more
convenient to downgrade this check to "info", which means
we'd still comment on it, but not reject a push. Site admins
can already do this via config, but we should ship sensible
defaults.
There are a few counterpoints to consider in favor of
keeping the check as an error:
- the first point above assumes that the config parsers for
the victim and the fsck process are equivalent. This is
pretty true now, but as time goes on will become less so.
Hosting sites are likely to upgrade their version of Git,
whereas vulnerable clients will be stagnant (if they did
upgrade, they'd cease to be vulnerable!). So in theory we
may see drift over time between what two config parsers
will accept.
In practice, this is probably OK. The config format is
pretty established at this point and shouldn't change a
lot. And the farther we get from the announcement of the
vulnerability, the less interesting this extra layer of
protection becomes. I.e., it was _most_ valuable on day
0, when everybody's client was still vulnerable and
hosting sites could protect people. But as time goes on
and people upgrade, the population of vulnerable clients
becomes smaller and smaller.
- In theory this could protect us from other
vulnerabilities in the future. E.g., .gitmodules are the
only way for a malicious repository to feed data to the
config parser, so this check could similarly protect
clients from a future (to-be-found) bug there.
But that's trading a hypothetical case for real-world
pain today. If we do find such a bug, the hosting site
would need to be updated to fix it, too. At which point
we could figure out whether it's possible to detect
_just_ the malicious case without hurting existing
broken-but-not-evil cases.
- Until recently, we hadn't made any restrictions on
.gitmodules content. So now in tightening that we're
hitting cases where certain things used to work, but
don't anymore. There's some moderate pain now. But as
time goes on, we'll see more (and more varied) cases that
will make tightening harder in the future. So there's
some argument for putting rules in place _now_, before
users grow more cases that violate them.
Again, this is trading pain now for hypothetical benefit
in the future. And if we try hard in the future to keep
our tightening to a minimum (i.e., rejecting true
maliciousness without hurting broken-but-not-evil repos),
then that reduces even the hypothetical benefit.
Considering both sets of arguments, it makes sense to loosen
this check for now.
Note that we have to tweak the test in t7415 since fsck will
no longer consider this a fatal error. But we still check
that it reports the warning, and that we don't get the
spurious error from the config code.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If there's a parsing error we'll already report it via the
usual fsck report() function (or not, if the user has asked
to skip this object or warning type). The error message from
the config parser just adds confusion. Let's suppress it.
Note that we didn't test this case at all, so I've added
coverage in t7415. We may end up toning down or removing
this fsck check in the future. So take this test as checking
what happens now with a focus on stderr, and not any
ironclad guarantee that we must detect and report parse
failures in the future.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"index-pack --strict" has been taught to make sure that it runs the
final object integrity checks after making the freshly indexed
packfile available to itself.
* jk/index-pack-maint:
index-pack: correct install_packed_git() args
index-pack: handle --strict checks of non-repo packs
prepare_commit_graft: treat non-repository as a noop
Commit 159e7b080b (fsck: detect gitmodules files,
2018-05-02) taught fsck to look at the content of
.gitmodules files. If the object turns out not to be a blob
at all, we just complain and punt on checking the content.
And since this was such an obvious and trivial code path, I
didn't even bother to add a test.
Except it _does_ do one non-trivial thing, which is call the
report() function, which wants us to pass a pointer to a
"struct object". Which we don't have (we have only a "struct
object_id"). So we erroneously pass a NULL object to
report(), which gets dereferenced and causes a segfault.
It seems like we could refactor report() to just take the
object_id itself. But we pass the object pointer along to
a callback function, and indeed this ends up in
builtin/fsck.c's objreport() which does want to look at
other parts of the object (like the type).
So instead, let's just use lookup_unknown_object() to get
the real "struct object", and pass that.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Early versions of the fsck .gitmodules detection code
actually required a tree to be at the root of a commit for
it to be checked for .gitmodules. What we ended up with in
159e7b080b (fsck: detect gitmodules files, 2018-05-02),
though, finds a .gitmodules file in _any_ tree (see that
commit for more discussion).
As a result, there's no need to create a commit in our
tests. Let's drop it in the name of simplicity. And since
that was the only thing referencing $tree, we can pull our
tree creation out of a command substitution.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 73c3f0f704 (index-pack: check .gitmodules files with
--strict, 2018-05-04) added a call to add_packed_git(), with
the intent that the newly-indexed objects would be available
to the process when we run fsck_finish(). But that's not
what add_packed_git() does. It only allocates the struct,
and you must install_packed_git() on the result. So that
call was effectively doing nothing (except leaking a
struct).
But wait, we passed all of the tests! Does that mean we
don't need the call at all?
For normal cases, no. When we run "index-pack --stdin"
inside a repository, we write the new pack into the object
directory. If fsck_finish() needs to access one of the new
objects, then our initial lookup will fail to find it, but
we'll follow up by running reprepare_packed_git() and
looking again. That logic was meant to handle somebody else
repacking simultaneously, but it ends up working for us
here.
But there is a case that does need this, that we were not
testing. You can run "git index-pack foo.pack" on any file,
even when it is not inside the object directory. Or you may
not even be in a repository at all! This case fails without
doing the proper install_packed_git() call.
We can make this work by adding the install call.
Note that we should be prepared to handle add_packed_git()
failing. We can just silently ignore this case, though. If
fsck_finish() later needs the objects and they're not
available, it will complain itself. And if it doesn't
(because we were able to resolve the whole fsck in the first
pass), then it actually isn't an interesting error at all.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We've recently forbidden .gitmodules to be a symlink in
verify_path(). And it's an easy way to circumvent our fsck
checks for .gitmodules content. So let's complain when we
see it.
Signed-off-by: Jeff King <peff@peff.net>
Now that the internal fsck code has all of the plumbing we
need, we can start checking incoming .gitmodules files.
Naively, it seems like we would just need to add a call to
fsck_finish() after we've processed all of the objects. And
that would be enough to cover the initial test included
here. But there are two extra bits:
1. We currently don't bother calling fsck_object() at all
for blobs, since it has traditionally been a noop. We'd
actually catch these blobs in fsck_finish() at the end,
but it's more efficient to check them when we already
have the object loaded in memory.
2. The second pass done by fsck_finish() needs to access
the objects, but we're actually indexing the pack in
this process. In theory we could give the fsck code a
special callback for accessing the in-pack data, but
it's actually quite tricky:
a. We don't have an internal efficient index mapping
oids to packfile offsets. We only generate it on
the fly as part of writing out the .idx file.
b. We'd still have to reconstruct deltas, which means
we'd basically have to replicate all of the
reading logic in packfile.c.
Instead, let's avoid running fsck_finish() until after
we've written out the .idx file, and then just add it
to our internal packed_git list.
This does mean that the objects are "in the repository"
before we finish our fsck checks. But unpack-objects
already exhibits this same behavior, and it's an
acceptable tradeoff here for the same reason: the
quarantine mechanism means that pushes will be
fully protected.
In addition to a basic push test in t7415, we add a sneaky
pack that reverses the usual object order in the pack,
requiring that index-pack access the tree and blob during
the "finish" step.
This already works for unpack-objects (since it will have
written out loose objects), but we'll check it with this
sneaky pack for good measure.
Signed-off-by: Jeff King <peff@peff.net>
As with the previous commit, we must call fsck's "finish"
function in order to catch any queued objects for
.gitmodules checks.
This second pass will be able to access any incoming
objects, because we will have exploded them to loose objects
by now.
This isn't quite ideal, because it means that bad objects
may have been written to the object database (and a
subsequent operation could then reference them, even if the
other side doesn't send the objects again). However, this is
sufficient when used with receive.fsckObjects, since those
loose objects will all be placed in a temporary quarantine
area that will get wiped if we find any problems.
Signed-off-by: Jeff King <peff@peff.net>
Now that the internal fsck code is capable of checking
.gitmodules files, we just need to teach its callers to use
the "finish" function to check any queued objects.
With this, we can now catch the malicious case in t7415 with
git-fsck.
Signed-off-by: Jeff King <peff@peff.net>
Submodule "names" come from the untrusted .gitmodules file,
but we blindly append them to $GIT_DIR/modules to create our
on-disk repo paths. This means you can do bad things by
putting "../" into the name (among other things).
Let's sanity-check these names to avoid building a path that
can be exploited. There are two main decisions:
1. What should the allowed syntax be?
It's tempting to reuse verify_path(), since submodule
names typically come from in-repo paths. But there are
two reasons not to:
a. It's technically more strict than what we need, as
we really care only about breaking out of the
$GIT_DIR/modules/ hierarchy. E.g., having a
submodule named "foo/.git" isn't actually
dangerous, and it's possible that somebody has
manually given such a funny name.
b. Since we'll eventually use this checking logic in
fsck to prevent downstream repositories, it should
be consistent across platforms. Because
verify_path() relies on is_dir_sep(), it wouldn't
block "foo\..\bar" on a non-Windows machine.
2. Where should we enforce it? These days most of the
.gitmodules reads go through submodule-config.c, so
I've put it there in the reading step. That should
cover all of the C code.
We also construct the name for "git submodule add"
inside the git-submodule.sh script. This is probably
not a big deal for security since the name is coming
from the user anyway, but it would be polite to remind
them if the name they pick is invalid (and we need to
expose the name-checker to the shell anyway for our
test scripts).
This patch issues a warning when reading .gitmodules
and just ignores the related config entry completely.
This will generally end up producing a sensible error,
as it works the same as a .gitmodules file which is
missing a submodule entry (so "submodule update" will
barf, but "git clone --recurse-submodules" will print
an error but not abort the clone.
There is one minor oddity, which is that we print the
warning once per malformed config key (since that's how
the config subsystem gives us the entries). So in the
new test, for example, the user would see three
warnings. That's OK, since the intent is that this case
should never come up outside of malicious repositories
(and then it might even benefit the user to see the
message multiple times).
Credit for finding this vulnerability and the proof of
concept from which the test script was adapted goes to
Etienne Stalmans.
Signed-off-by: Jeff King <peff@peff.net>
