]> git.apps.os.sepia.ceph.com Git - fscrypt.git/log
fscrypt.git
4 years agoREADME: clarify how restoring /.fscrypt directory works
Eric Biggers [Tue, 14 Sep 2021 18:07:42 +0000 (11:07 -0700)]
README: clarify how restoring /.fscrypt directory works

Update https://github.com/google/fscrypt/issues/115

4 years agoMerge pull request #310 from ebiggers/readme-updates
Joseph Richey [Tue, 14 Sep 2021 03:16:07 +0000 (20:16 -0700)]
Merge pull request #310 from ebiggers/readme-updates

Readme updates

4 years agoREADME: add section about encrypting existing files
Eric Biggers [Mon, 13 Sep 2021 23:23:50 +0000 (16:23 -0700)]
README: add section about encrypting existing files

4 years agoREADME: consistently format section titles
Eric Biggers [Mon, 13 Sep 2021 21:45:34 +0000 (14:45 -0700)]
README: consistently format section titles

Capitalize the first word only, and don't use periods.

4 years agoREADME: add section about backup, restore, and recovery
Eric Biggers [Mon, 13 Sep 2021 19:40:14 +0000 (12:40 -0700)]
README: add section about backup, restore, and recovery

Resolves https://github.com/google/fscrypt/issues/51
Resolves https://github.com/google/fscrypt/issues/115

4 years agoREADME: remove note about stability
Eric Biggers [Mon, 13 Sep 2021 19:40:14 +0000 (12:40 -0700)]
README: remove note about stability

A lot of people are already using fscrypt, so in practice we haven't
been breaking backwards compatibility and aren't going to.  Just remove
the scary-sounding "Note about stability".

4 years agoREADME: remove note about planned commands
Eric Biggers [Mon, 13 Sep 2021 19:40:14 +0000 (12:40 -0700)]
README: remove note about planned commands

These would still be nice to add.  However, the mention of them in the
README is misleading because people reading it might come away with the
impression that there is currently no way to back up fscrypt metadata or
to recover directories -- which isn't true.  (The fscrypt metadata is
just a directory which can be backed up like any other directory.  And
'fscrypt encrypt' already offers to generate a recovery passphrase when
the directory and protector are on different filesystems.)

Just remove this note; it doesn't really add any value.

4 years agoMerge pull request #309 from ebiggers/troubleshooting
Joseph Richey [Wed, 1 Sep 2021 02:14:08 +0000 (19:14 -0700)]
Merge pull request #309 from ebiggers/troubleshooting

Updates to the troubleshooting documentation

4 years agoREADME: document symlink size bug
Eric Biggers [Tue, 31 Aug 2021 23:47:00 +0000 (16:47 -0700)]
README: document symlink size bug

Update https://github.com/google/fscrypt/issues/305

4 years agoREADME: documentation tweaks
Eric Biggers [Thu, 26 Aug 2021 22:54:06 +0000 (15:54 -0700)]
README: documentation tweaks

Clarify some of the troubleshooting documentation.

4 years agocli-tests: fix failure with latest bash version
Eric Biggers [Fri, 16 Jul 2021 22:59:26 +0000 (17:59 -0500)]
cli-tests: fix failure with latest bash version

4 years agoREADME: add troubleshooting section for ENOPKG error
Eric Biggers [Fri, 16 Jul 2021 22:31:58 +0000 (17:31 -0500)]
README: add troubleshooting section for ENOPKG error

Inspired by https://reddit.com/r/linuxquestions/comments/n75dv4

4 years agoREADME: add troubleshooting section for keyring linkage issue
Eric Biggers [Sun, 27 Jun 2021 21:56:15 +0000 (14:56 -0700)]
README: add troubleshooting section for keyring linkage issue

Fixes https://github.com/google/fscrypt/issues/194
Update https://github.com/google/fscrypt/issues/100

4 years agocmd/fscrypt: fix detection of GRUB installation
Eric Biggers [Sun, 27 Jun 2021 20:13:10 +0000 (13:13 -0700)]
cmd/fscrypt: fix detection of GRUB installation

Fix the GRUB detection logic to take into account that
MOUNTPOINT/boot/grub might not be on the same filesystem as MOUNTPOINT,
due to MOUNTPOINT/boot being another mountpoint.  The warning is only
appropriate when GRUB is installed on the same filesystem that
encryption is going to be enabled on.

4 years agoREADME: improve troubleshooting tips for unlocked encrypted files
Eric Biggers [Thu, 10 Jun 2021 04:21:22 +0000 (21:21 -0700)]
README: improve troubleshooting tips for unlocked encrypted files

Rename the troubleshooting section "Can't log in with ssh even when
user's encrypted home directory is unlocked" to the more general "Some
processes can't access unlocked encrypted files", and rewrite it to
provide clearer directions for how to fix the problem by upgrading
encrypted directories to policy version 2.

Also add a related section "Users can access other users' unlocked
encrypted files" which covers the reverse "issue", i.e. people expecting
some processes to *not* be able to access unlocked encrypted files.

Fixes https://github.com/google/fscrypt/issues/248

4 years agoREADME: improve troubleshooting section for login protector not in sync
Eric Biggers [Tue, 8 Jun 2021 23:23:04 +0000 (16:23 -0700)]
README: improve troubleshooting section for login protector not in sync

Update https://github.com/google/fscrypt/issues/273

4 years agoMerge pull request #292 from josephlr/mem
Joseph Richey [Tue, 25 May 2021 05:18:42 +0000 (22:18 -0700)]
Merge pull request #292 from josephlr/mem

Only use 1/8 of the system RAM and run the Garbage Collector in the timing loop

4 years agoRun the Garbage Collector in the timing loop
Joe Richey [Mon, 24 May 2021 10:42:01 +0000 (03:42 -0700)]
Run the Garbage Collector in the timing loop

Running `crypto.PassphraseHash` in a loop allocates a lot of memory.
Golang is not always prudent about collecting the garbage from previous
runs, resulting in a OOM error on memory-pressured systems.

With a `maxMemoryBytes` of 128 MiB, this change reduces the maximum
resident memory for `fscrypt setup` to 141 MiB (was perviously 405 MiB)

Signed-off-by: Joe Richey <joerichey@google.com>
4 years agoOnly use 1/8 of the system RAM
Joe Richey [Mon, 24 May 2021 10:41:16 +0000 (03:41 -0700)]
Only use 1/8 of the system RAM

On systems with high memory pressure, using half of the entire RAM for
hashing can result in fscrypt getting OOM killed.

Signed-off-by: Joe Richey <joerichey@google.com>
4 years agofilesystem: skip TestHaveReadAccessTo() when running as root
Eric Biggers [Wed, 12 May 2021 20:28:30 +0000 (13:28 -0700)]
filesystem: skip TestHaveReadAccessTo() when running as root

Root can read all files, so this test fails when running as root.
Skip it instead.

Resolves https://github.com/google/fscrypt/issues/288

4 years agoSpecify -buildmode=c-shared after GO_FLAGS rather than before
Eric Biggers [Thu, 6 May 2021 05:09:26 +0000 (22:09 -0700)]
Specify -buildmode=c-shared after GO_FLAGS rather than before

When building pam_fscrypt.so, specify -buildmode=c-shared after
$(GO_FLAGS) so that it overrides any user-specified buildmode.

This is needed to allow -buildmode=pie to be specified in GO_FLAGS if
the packager wants to build fscrypt as a position-independent executable
(e.g. following https://wiki.archlinux.org/title/Go_package_guidelines).
Previously, trying to do this caused pam_fscrypt.so to be incorrectly
built as an executable rather than as a shared library.

4 years agocmd/fscrypt: fix word mismatch "protector" => "policy"
Gibeom Gwon [Tue, 27 Apr 2021 10:49:01 +0000 (19:49 +0900)]
cmd/fscrypt: fix word mismatch "protector" => "policy"

Fix word mismatch in usage and description of metadata create policy
command.

4 years agocmd/fscrypt: use golang.org/x/term
Tobias Klauser [Thu, 22 Apr 2021 11:53:56 +0000 (13:53 +0200)]
cmd/fscrypt: use golang.org/x/term

The golang.org/x/crypto/ssh/terminal package is deprecated and merely a
wrapper around golang.org/x/term. Thus, use the latter directly.

4 years agoMakefile: Optionally avoid installation of Ubuntu-specific PAM config
Florian Schmaus [Sat, 17 Apr 2021 16:19:59 +0000 (18:19 +0200)]
Makefile: Optionally avoid installation of Ubuntu-specific PAM config

This allows non Ubuntu distributions to opt out from the installation
of Ubuntu-specific PAM files.

4 years agoRelease version v0.3.0 (#282)
Eric Biggers [Wed, 31 Mar 2021 23:19:39 +0000 (16:19 -0700)]
Release version v0.3.0 (#282)

4 years agoMerge pull request #283 from google/coverage
Joseph Richey [Wed, 31 Mar 2021 23:08:48 +0000 (16:08 -0700)]
Merge pull request #283 from google/coverage

Stop generating and uploading coverage in CI

4 years agoOnly run CI on master branch and PRs to master
Joe Richey [Wed, 31 Mar 2021 23:00:16 +0000 (16:00 -0700)]
Only run CI on master branch and PRs to master

This avoids duplicate CI checks

Signed-off-by: Joe Richey <joerichey@google.com>
4 years agoStop generating and uploading coverage in CI
Joe Richey [Wed, 31 Mar 2021 22:46:18 +0000 (15:46 -0700)]
Stop generating and uploading coverage in CI

This is currently broken, and we don't really use the findings.

Signed-off-by: Joe Richey <joerichey@google.com>
4 years agoMerge pull request #281 from ebiggers/pam_fscrypt-updates
Joseph Richey [Tue, 9 Mar 2021 09:27:42 +0000 (01:27 -0800)]
Merge pull request #281 from ebiggers/pam_fscrypt-updates

pam_fscrypt: eliminate unnecessary options and improve documentation

4 years agoREADME: improve PAM configuration documentation (again)
Eric Biggers [Mon, 8 Mar 2021 23:20:08 +0000 (15:20 -0800)]
README: improve PAM configuration documentation (again)

Make some more corrections:

- pam-config-framework isn't actually Ubuntu-specific but actually
  applies to Debian and any Debian derivative.

- The pam-config-framework file is indeed installed by `make install`,
  just not into the correct location.

- On Debian (and Debian derivatives), the PAM configuration isn't
  actually part of the 'fscrypt' package but rather 'libpam-fscrypt'.

- Clarify where to add the pam_fscrypt.so session hook.

4 years agoREADME: make it clear that pam_fscrypt also handles locking
Eric Biggers [Mon, 8 Mar 2021 23:20:08 +0000 (15:20 -0800)]
README: make it clear that pam_fscrypt also handles locking

There are several mentions of pam_fscrypt handling unlocking
directories.  Make sure to mention locking alongside this.

4 years agopam_fscrypt: make "lock_policies" the default behavior
Eric Biggers [Mon, 8 Mar 2021 23:20:08 +0000 (15:20 -0800)]
pam_fscrypt: make "lock_policies" the default behavior

All pam_fscrypt configuration guides that I'm aware of say to use the
"lock_policies" option for the pam_fscrypt.so session hook.  The
Debian/Ubuntu pam-config-framework config file has it too.

Make locking the default behavior, since this is what everyone wants.

Existing configuration files that contain the "lock_policies" option
will continue to work, but that option won't do anything anymore.

(We could add an option "unlock_only" to restore the old default
behavior, but it's not clear that it would be useful.  So for
simplicity, leave it out for now.)

4 years agopam_fscrypt: decide cache dropping behavior automatically
Eric Biggers [Mon, 8 Mar 2021 23:20:08 +0000 (15:20 -0800)]
pam_fscrypt: decide cache dropping behavior automatically

Configuring whether pam_fscrypt drops caches or not isn't really
something the user should have to do, and it's also irrelevant for v2
encryption policies (the default on newer systems).  It's better to have
pam_fscrypt automatically decide whether it needs to drop caches or not.

Do this by making pam_fscrypt check whether any encryption policy keys
are being removed from a user keyring (rather than from a filesystem
keyring).  If so, it drops caches; otherwise it doesn't.  This
supersedes the "drop_caches" option, which won't do anything anymore.

4 years agopam_fscrypt/config: prioritise over other session modules
Robert McQueen [Wed, 3 Mar 2021 11:34:55 +0000 (11:34 +0000)]
pam_fscrypt/config: prioritise over other session modules

Services launched by systemd user sessions on Debian / Ubuntu systems
are often not able to access the home directory, because there is no
guarantee / requirement that pam_fscrypt is sequenced before
pam_systemd.

Although this pam-config mechanism is Debian-specific, the config file
is provided here upstream and unmodified in Debian. Raising the
priority here so that it's always ordered ahead of pam_systemd will
solve issues such as https://github.com/google/fscrypt/issues/270,
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=964951 and
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1889416.

After a survey of pam-config files available in Debian bullseye, the
value of 100 was chosen as it appears after most other plugins that
could be involved in more explicit homedir configuration (eg pam_mount
at 128) but before those which seem unlikely to work without a home
directory (eg pam_ssh at 64).

4 years agoci.yml: always run apt-get update
Eric Biggers [Wed, 3 Feb 2021 00:14:30 +0000 (16:14 -0800)]
ci.yml: always run apt-get update

In GitHub Workflows, apparently running 'apt-get update' before 'apt-get
install' is sometimes needed, and it doesn't hurt to always do it.

4 years agoAvoid using the word "whitelist"
Eric Biggers [Mon, 25 Jan 2021 19:12:01 +0000 (11:12 -0800)]
Avoid using the word "whitelist"

4 years agocmd/fscrypt: fix missing protector error format
Alastair Hughes [Fri, 15 Jan 2021 10:08:54 +0000 (23:08 +1300)]
cmd/fscrypt: fix missing protector error format

Update #272

4 years agobash-completion: add completion script
Henry-Joseph Audéoud [Fri, 30 Oct 2020 10:41:52 +0000 (11:41 +0100)]
bash-completion: add completion script

4 years agoREADME: Fix badge image
Joe Richey [Thu, 26 Nov 2020 09:25:12 +0000 (01:25 -0800)]
README: Fix badge image

Workflow names are case-sensitive

Signed-off-by: Joe Richey <joerichey@google.com>
4 years agoSwitch from Travis CI to GitHub Actions
Eric Biggers [Sat, 21 Nov 2020 23:29:26 +0000 (15:29 -0800)]
Switch from Travis CI to GitHub Actions

travis-ci.org is being shut down, so switch to GitHub Actions.

It should be mostly equivalent, but I did drop functionality in a couple
cases:

- Publishing release binaries.  I don't think providing Linux binaries
  is useful, since people build their own anyway.  So I left this out.

- Build and testing on ppc64le.  GitHub Actions only natively supports
  x86.  I tried uraimo/run-on-arch-action, which uses Docker and QEMU
  user-mode emulation, but the fscrypt tests can't be run because
  QEMU user-mode emulation doesn't support all the needed system calls.

4 years agocli-tests: force processes spawned by 'expect' to have 80 column-output
Eric Biggers [Sat, 21 Nov 2020 23:29:26 +0000 (15:29 -0800)]
cli-tests: force processes spawned by 'expect' to have 80 column-output

Otherwise the cli tests fail when executed from GitHub Actions.

4 years agoREADME.md: add table of contents
Eric Biggers [Tue, 10 Nov 2020 04:43:48 +0000 (20:43 -0800)]
README.md: add table of contents

4 years agotravis.yml: upgrade to Ubuntu 20.04 and enable cli tests
Eric Biggers [Sun, 8 Nov 2020 04:30:51 +0000 (20:30 -0800)]
travis.yml: upgrade to Ubuntu 20.04 and enable cli tests

Now that Travis CI supports a version of Ubuntu that has a kernel that
supports v2 encryption policies, upgrade to it and enable the cli tests.

4 years agocmd/fscrypt: fix race condition in getPassphraseKey()
Eric Biggers [Sun, 8 Nov 2020 04:30:51 +0000 (20:30 -0800)]
cmd/fscrypt: fix race condition in getPassphraseKey()

Set the terminal to raw mode *before* printing the prompt.
Otherwise the user (or the automated test) might enter the
passphrase before the terminal gets put into raw mode.

This is needed for some of the CLI tests to pass reliably in Travis CI.

4 years agocli-tests/common.sh: fix _user_do()
Eric Biggers [Sun, 8 Nov 2020 04:30:51 +0000 (20:30 -0800)]
cli-tests/common.sh: fix _user_do()

Apparently, on some distros 'su' doesn't preserve $PATH.  So, manually
export it in the command.  Also, ensure that the shell stays as bash.

This is needed for some of the CLI tests to pass in Travis CI.

4 years agocmd/fscrypt: fix isDirUnlockedHeuristic() on latest kernels
Eric Biggers [Sat, 7 Nov 2020 22:20:45 +0000 (14:20 -0800)]
cmd/fscrypt: fix isDirUnlockedHeuristic() on latest kernels

On an "incompletely locked" directory, isDirUnlockedHeuristic() is
supposed to return true, but on Linux v5.10-rc1 and later it returns
false since now creating a subdirectory fails rather than succeeds.
This change was intentional, so make isDirUnlockedHeuristic() apply a
second heuristic too: also return true if any filenames in the directory
don't appear to be valid no-key names.

This fixes cli-tests/t_v1_encrypt on Linux v5.10-rc1 and later.

4 years agoTravis-ci: added support for ppc64le (#257)
sanjaymsh [Thu, 15 Oct 2020 12:27:25 +0000 (17:57 +0530)]
Travis-ci: added support for ppc64le (#257)

5 years agoFix nil error issue, Resolves https://github.com/google/fscrypt/issues/242
bitcodr [Fri, 7 Aug 2020 05:34:34 +0000 (10:04 +0430)]
Fix nil error issue, Resolves https://github.com/google/fscrypt/issues/242

5 years agoREADME.md: recommend 'sudo make install PREFIX=/usr' on Ubuntu (#244)
Eric Biggers [Fri, 7 Aug 2020 23:37:05 +0000 (16:37 -0700)]
README.md: recommend 'sudo make install PREFIX=/usr' on Ubuntu (#244)

Ubuntu's PAM configuration framework only recognizes files in /usr, not
/usr/local.  So for installs from source, unfortunately we have to
recommend installing to /usr, despite this not being conventional.

Resolves https://github.com/google/fscrypt/issues/240

5 years agoRelease version v0.2.9 (#238)
Eric Biggers [Sat, 13 Jun 2020 17:43:29 +0000 (10:43 -0700)]
Release version v0.2.9 (#238)

5 years agoMerge pull request #237 from ebiggers/t_v1_policy_fix
Eric Biggers [Sat, 13 Jun 2020 17:27:57 +0000 (10:27 -0700)]
Merge pull request #237 from ebiggers/t_v1_policy_fix

Adjust status message for v1 policies unlocked by another user and fix cli-tests/t_v1_policy

5 years agocmd/fscrypt: adjust status message for v1-encrypted dirs
Eric Biggers [Sat, 13 Jun 2020 17:06:15 +0000 (10:06 -0700)]
cmd/fscrypt: adjust status message for v1-encrypted dirs

When 'fscrypt status DIR' detects that a v1-encrypted directory is still
usable but its key seems to be absent, it shows the status as
"Unlocked: Partially (incompletely locked)".  But actually it can also
be the case that the directory is unlocked by another user.  Adjust the
status message accordingly.

This commit also fixes cli-tests/t_v1_policy.

5 years agocli-tests/t_v1_policy: clean up user keyrings at end of test
Eric Biggers [Sat, 13 Jun 2020 17:06:15 +0000 (10:06 -0700)]
cli-tests/t_v1_policy: clean up user keyrings at end of test

The test user's user keyring is still linked into root's user keyring at
the end of the test.  This is making the test flaky, as there is a
failure that only occurs the first time it is run.  Fix the test to
restore the initial state.  This makes it consistently fail (to be fixed
by the next commit).

5 years agoMerge pull request #235 from ebiggers/32bit-fix
Eric Biggers [Wed, 3 Jun 2020 01:02:08 +0000 (18:02 -0700)]
Merge pull request #235 from ebiggers/32bit-fix

* cmd/fscrypt: fix 32-bit build
* travis.yml: build 32-bit binary

5 years agocmd/fscrypt: fix 32-bit build
Eric Biggers [Wed, 3 Jun 2020 00:17:54 +0000 (17:17 -0700)]
cmd/fscrypt: fix 32-bit build

statfs.Bsize actually has platform-dependent type, despite the Go
documentation listing it as int64.  Fix the build for 32-bit platforms
by casting it to int64.

Resolves https://github.com/google/fscrypt/issues/233

5 years agotravis.yml: build 32-bit binary
Eric Biggers [Wed, 3 Jun 2020 00:09:21 +0000 (17:09 -0700)]
travis.yml: build 32-bit binary

Don't let people check in code that breaks 32-bit builds.

Update https://github.com/google/fscrypt/issues/233

5 years agoRelease version v0.2.8 (#230)
Eric Biggers [Wed, 20 May 2020 10:36:16 +0000 (03:36 -0700)]
Release version v0.2.8 (#230)

5 years agocmd/fscrypt: fix up path formatting in ErrDirNotEmpty suggestion (#229)
Eric Biggers [Fri, 15 May 2020 02:48:23 +0000 (19:48 -0700)]
cmd/fscrypt: fix up path formatting in ErrDirNotEmpty suggestion (#229)

Use %q, in case the paths contain whitespace.  Also clean the directory
path to remove trailing slashes before appending the ".new" suffix.

5 years agocmd/fscrypt: link to guide when interactively creating login protector (#225)
Eric Biggers [Thu, 14 May 2020 02:14:13 +0000 (19:14 -0700)]
cmd/fscrypt: link to guide when interactively creating login protector (#225)

Update https://github.com/google/fscrypt/issues/220

5 years agoREADME.md: try to disambiguate "fscrypt" (#226)
Eric Biggers [Thu, 14 May 2020 02:11:28 +0000 (19:11 -0700)]
README.md: try to disambiguate "fscrypt" (#226)

Explicitly mention that "fscrypt" here means the userspace tool, not the
kernel part.  Also write `fscrypt` in code font to emphasize this.

5 years agoREADME.md: further improve the "encryption not enabled" section (#227)
Eric Biggers [Thu, 14 May 2020 02:10:18 +0000 (19:10 -0700)]
README.md: further improve the "encryption not enabled" section (#227)

5 years agoREADME.md: fix typo "nrounds" => "rounds" (#228)
Eric Biggers [Thu, 14 May 2020 01:53:56 +0000 (18:53 -0700)]
README.md: fix typo "nrounds" => "rounds" (#228)

Update https://github.com/google/fscrypt/issues/220

5 years agoREADME.md: improve documentation for login protectors (#223)
Eric Biggers [Tue, 12 May 2020 16:45:42 +0000 (09:45 -0700)]
README.md: improve documentation for login protectors (#223)

Update https://github.com/google/fscrypt/issues/220

5 years agoDeclare instead of define variable in header file. (#224)
Shuai Wang [Sun, 10 May 2020 21:22:54 +0000 (14:22 -0700)]
Declare instead of define variable in header file. (#224)

This fixes link error with gcc 10 which defaults to -fno-common

5 years agoMerge pull request #219 from ebiggers/improve-errors
Eric Biggers [Sat, 9 May 2020 22:27:24 +0000 (15:27 -0700)]
Merge pull request #219 from ebiggers/improve-errors

Improve error messages and suggestions

5 years agocmd/fscrypt: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
cmd/fscrypt: improve errors

In checkEncryptable(), check whether the directory is already encrypted
before checking whether it's empty.

Also improve the error message for when a directory is nonempty.

Finally, translate keyring.ErrKeyAddedByOtherUsers and
keyring.ErrKeyFilesOpen into errors which include the directory.

5 years agocmd/fscrypt: remove ErrMaxPassphrase
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
cmd/fscrypt: remove ErrMaxPassphrase

This isn't actually a valid error since crypto.NewKeyFromReader()
handles re-allocating the buffer to a larger size if it fills up.

5 years agofilesystem: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
filesystem: improve errors

Introduce filesystem.ErrEncryptionNotEnabled and
filesystem.ErrEncryptionNotSupported which include the Mount as context,
and translate the corresponding metadata/ errors into them.  Then make
these errors show much better suggestions.

Also replace lots of other filesystem/ errors with either custom types
or with unnamed one-off errors that include more context.  Fix backwards
wrapping in lots of cases.

Finally, don't include the mountpoint in places where it's not useful,
like OS-level errors that already include the path.

5 years agometadata: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
metadata: improve errors

ErrBadOwners:
Rename to ErrDirectoryNotOwned for clarity, move it from
cmd/fscrypt/ to metadata/ where it better belongs, and improve
the message.

ErrEncrypted:
Rename to ErrAlreadyEncrypted for clarity, and include the path.

ErrNotEncrypted:
Include the path.

ErrBadEncryptionOptions:
Include the path and bad options.

ErrEncryptionNotSupported:
ErrEncryptionNotEnabled:
Don't wrap with "get encryption policy %s", in preparation for
wrapping these with filesystem-level context instead.

Also avoid mixing together the error handling for the "get policy" and
"set policy" ioctls.  Make it very clear how we're handling the errors
from each ioctl.

5 years agokeyring: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
keyring: improve errors

ErrAccessUserKeyring:
Include the user, and fix the backwards wrapping.

ErrSessionUserKeyring:
Include the user.

ErrKeyAdd:
ErrKeyRemove:
ErrKeySearch:
ErrLinkUserKeyring:
Replace these with one-off unnamed errors because they are
never checked for, and this makes it easier for the callers to
provide better messages, e.g. fixing the backwards wrapping.

5 years agocrypto: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
crypto: improve errors

ErrKeyLock:
Rename to ErrMlockUlimit for clarity.

ErrGetrandomFail:
ErrKeyAlloc:
ErrKeyFree:
ErrNegativeLength:
Replace these with one-off unnamed errors because these were all
returned in only one place and were never checked for.  Also
these were all either wrapped backwards or discarded an
underlying error, so fix that too.

5 years agoactions/policy: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
actions/policy: improve errors

ErrMissingPolicyMetadata:
Include the mount, directory path, and metadata path.  Also move
the explanation into actions/ since it doesn't refer to any CLI
command.

ErrPolicyMetadataMismatch:
Include a lot more information.  Also start checking for
consistency of the policy key descriptors, not just the
encryption options.  Add a test for this.

ErrDifferentFilesystem:
Include the mountpoints.

ErrOnlyProtector:
Clarify the message and include the protector descriptor.

ErrAlreadyProtected:
ErrNotProtected:
Include the policy and protector descriptors.

ErrAccessDeniedPossiblyV2:
Make it slightly clearer what failed.  Also move the explanation
into actions/ since it doesn't refer to any CLI command.

5 years agoactions/protector: improve errors
Eric Biggers [Sat, 9 May 2020 21:52:07 +0000 (14:52 -0700)]
actions/protector: improve errors

ErrProtectorName:
Rename to ErrLoginProtectorName for clarity, and include the
name and user.

ErrMissingProtectorName:
Include the correct protector source.

ErrDuplicateName:
Rename to ErrProtectorNameExists for clarity, and remove a level
of wrapping by including the name directly.

ErrDuplicateUID:
Rename to ErrLoginProtectorExists for clarity, and remove a
level of wrapping by including the user directly.

5 years agoactions/config: improve config file related errors
Eric Biggers [Sat, 9 May 2020 21:52:06 +0000 (14:52 -0700)]
actions/config: improve config file related errors

ErrBadConfig:
Fix backwards wrapping, include the bad config, and make it
clear that this is an internal error.

ErrBadConfigFile:
Fix backwards wrapping, include the config file location, and
adjust the suggestion slightly.

ErrConfigFileExists:
Include the config file location.

ErrNoConfigFile:
Include the config file location, and adjust the suggestion
slightly.

5 years agocmd/fscrypt: make wrapText() support code blocks
Eric Biggers [Sat, 9 May 2020 21:52:06 +0000 (14:52 -0700)]
cmd/fscrypt: make wrapText() support code blocks

Allow the input text to contain "code blocks" denoted by lines beginning
with ">", e.g.:

    Foo bar baz:

    > echo foo
    > echo bar

Instead of squashing these lines together, preserve the line breaks
between them and add indentation, e.g.:

    Foo bar baz:

        echo foo
        echo bar

5 years agoMerge pull request #217 from ebiggers/detect-incomplete-v1-locking
Eric Biggers [Sat, 9 May 2020 22:21:07 +0000 (15:21 -0700)]
Merge pull request #217 from ebiggers/detect-incomplete-v1-locking

Try to detect incomplete locking of v1-encrypted directory

5 years agoTry to detect incomplete locking of v1-encrypted directory
Eric Biggers [Sat, 9 May 2020 21:17:17 +0000 (14:17 -0700)]
Try to detect incomplete locking of v1-encrypted directory

'fscrypt lock' on a v1-encrypted directory doesn't warn about in-use
files, as the kernel doesn't provide a way to easily detect it.

Instead, implement a heuristic where we check whether a subdirectory can
be created.  If yes, then the directory must not be fully locked.

Make both 'fscrypt lock' and 'fscrypt status' use this heuristic.

Resolves https://github.com/google/fscrypt/issues/215

5 years agokeyring/user_keyring: switch to KEYCTL_UNLINK
Eric Biggers [Sat, 9 May 2020 21:17:17 +0000 (14:17 -0700)]
keyring/user_keyring: switch to KEYCTL_UNLINK

KEYCTL_INVALIDATE has complicated semantics: it doesn't remove the key
from the keyring right away but rather marks it as being invalidated,
and then removes it asynchronously.  This nondeterministically breaks
the heuristic I'm implementing to detect v1-encrypted directories being
incompletely locked.

Instead, switch to KEYCTL_UNLINK, which has simpler semantics.

Note that Android uses KEYCTL_UNLINK too.

5 years agoMerge pull request #218 from ebiggers/cli-tests
Eric Biggers [Sat, 9 May 2020 22:15:12 +0000 (15:15 -0700)]
Merge pull request #218 from ebiggers/cli-tests

Add tests for command-line interface

Add tests that directly test the fscrypt command-line tool.

See cli-tests/README.md for information about the test framework.

The following test scripts are included:

* t_change_passphrase
* t_encrypt_custom
* t_encrypt_login
* t_encrypt_raw_key
* t_encrypt
* t_lock
* t_not_enabled
* t_not_supported
* t_passphrase_hashing
* t_setup
* t_status
* t_unlock
* t_v1_policy_fs_keyring
* t_v1_policy

Unfortunately, we can't actually make Travis CI run these tests yet because they need kernel v5.4 or later, and Travis CI doesn't support an Ubuntu version that has that yet. But for now, they can be run manually using make cli-test.

5 years agocli-tests: add t_v1_policy_fs_keyring
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_v1_policy_fs_keyring

Test using v1 encryption policies (deprecated) with
use_fs_keyring_for_v1_policies = true.

5 years agocli-tests: add t_v1_policy
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_v1_policy

Test using v1 encryption policies (deprecated).

5 years agocli-tests: add t_unlock
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_unlock

Test unlocking a directory.

5 years agocli-tests: add t_status
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_status

Test getting global, filesystem, and unencrypted directory status when
the filesystem is or isn't set up for fscrypt.

5 years agocli-tests: add t_setup
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_setup

Test 'fscrypt setup'.

5 years agocli-tests: add t_passphrase_hashing
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_passphrase_hashing

Test that the passphrase hashing seems to take long enough.

5 years agocli-tests: add t_not_supported
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_not_supported

Test that fscrypt fails when the filesystem doesn't support encryption.

5 years agocli-tests: add t_not_enabled
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_not_enabled

Test that fscrypt fails when the filesystem doesn't have the encrypt
feature enabled.  Then test enabling it.

5 years agocli-tests: add t_lock
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_lock

Test locking a directory.

5 years agocli-tests: add t_encrypt_raw_key
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_encrypt_raw_key

Test encrypting a directory using a raw_key protector.

5 years agocli-tests: add t_encrypt_login
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_encrypt_login

Test encrypting a directory using a login (pam_passphrase) protector.

5 years agocli-tests: add t_encrypt_custom
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_encrypt_custom

Test encrypting a directory using a custom_passphrase protector.

5 years agocli-tests: add t_encrypt
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_encrypt

Add general tests for 'fscrypt encrypt'.  For protector-specific tests,
see t_encrypt_custom, t_encrypt_login, and t_encrypt_raw_key.

5 years agocli-tests: add t_change_passphrase
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cli-tests: add t_change_passphrase

Test changing the passphrase of a custom_passphrase protector.

5 years agoWire up shellcheck of cli-tests/*.sh to 'make lint'
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
Wire up shellcheck of cli-tests/*.sh to 'make lint'

Enforce that all the cli-test scripts pass 'shellcheck'.

5 years agoAdd cli-tests framework
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
Add cli-tests framework

Add a framework for writing automated tests of the fscrypt command-line
tool.  See cli-tests/README.md for details.

5 years agocmd/fscrypt: add FSCRYPT_CONSISTENT_OUTPUT environmental variable
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cmd/fscrypt: add FSCRYPT_CONSISTENT_OUTPUT environmental variable

Allow setting FSCRYPT_CONSISTENT_OUTPUT=1 in the environment to cause
policies and protectors to sorted by last modification time.  The CLI
tests need this to make the output of 'fscrypt' ordered in a consistent
way with regard to the operations performed.

5 years agocmd/fscrypt: add FSCRYPT_ROOT_MNT environmental variable
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cmd/fscrypt: add FSCRYPT_ROOT_MNT environmental variable

Allow overriding the mountpoint where login protectors are stored by
setting the FSCRYPT_ROOT_MNT environmental variable.  The CLI tests need
this to avoid touching the real "/".

5 years agocmd/fscrypt: add FSCRYPT_CONF environmental variable
Eric Biggers [Sat, 9 May 2020 21:04:47 +0000 (14:04 -0700)]
cmd/fscrypt: add FSCRYPT_CONF environmental variable

Allow overriding the location of fscrypt.conf by setting the
FSCRYPT_CONF environmental variable.  The CLI tests need this to avoid
touching the real /etc/fscrypt.conf.

5 years agotravis.yml: stop overriding e2fsprogs version (#222)
Eric Biggers [Sat, 9 May 2020 19:54:37 +0000 (12:54 -0700)]
travis.yml: stop overriding e2fsprogs version (#222)

There's no longer a need to override the Ubuntu version that the
Travis CI builds install e2fsprogs from, since we now use
"dist: bionic", and e2fsprogs in Bionic supports encryption.

5 years agokeyring: cast FS_IOC_REMOVE_ENCRYPTION_KEY to uintptr (#221)
Filip Stanis [Tue, 5 May 2020 01:48:26 +0000 (02:48 +0100)]
keyring: cast FS_IOC_REMOVE_ENCRYPTION_KEY to uintptr (#221)

Since v0.2.6, fscrypt only builds for 64-bit systems.  E.g. trying to
build on Raspbian fails with the following error:

  $ go get github.com/google/fscrypt/cmd/fscrypt
  # github.com/google/fscrypt/keyring
  go/src/github.com/google/fscrypt/keyring/fs_keyring.go:231:6: constant 3225445912 overflows int
  go/src/github.com/google/fscrypt/keyring/fs_keyring.go:235:7: constant 3225445913 overflows int

Fix it by making the 'ioc' variable have type uintptr.

[EB - removed the later cast to uintptr that became unnecessary, and
 added explanation to commit message.]