This is already on for all Pixel devices with no problems observed.
If this causes issues with a specific device (e.g. vendor apps being
unable to access their data) it can be temporarily disabled by adding
PRODUCT_PROPERTY_OVERRIDES += ro.vold.level_from_user=0
to the device.mk file. Please file a bug if that happens.
Bug: 141677108
Test: presubmits
Change-Id: Ic9da534f1a5f4c9e3bd62ea5c09a3b11ebcb33e7
Merged-In: Ic9da534f1a5f4c9e3bd62ea5c09a3b11ebcb33e7
(cherry picked from commit 763393644a)
Remove the error-prone 'keepOld' parameter, and instead make begin()
(renamed to BeginKeymasterOp()) do all the key upgrade handling.
Don't handle /data and /metadata differently anymore. Previously, when
a checkpoint is active, key blob files were replaced on /data
immediately; only the actual Keymaster key deletion was delayed until
checkpoint commit. But it's easier to just delay the key blob file
replacement too, as we have to implement that for /metadata anyway.
Also be more vigilant about deleting any leftover upgraded keys.
Test: Tested on bramble using an OTA rvc-d1-release => master. In OTA
success case, verified via logcat that the keys were upgraded and
then were committed after the boot succeeded. In OTA failure
case, verified that the device still boots -- i.e., the old keys
weren't lost. Verified that in either case, no
keymaster_key_blob_upgraded files were left over. Finally, also
tried 'pm create-user' and 'pm remove-user' and verified via
logcat that the Keymaster keys still get deleted.
Change-Id: Ic9c3e63e0bcae0c608fc79050ca4a1676b3852ee
Commit 77df7f207d / http://aosp/1217657 ("Refactor to use
EncryptionPolicy everywhere we used to use raw_ref") unintentionally
made fscrypt_initialize_systemwide_keys() start specifying keepOld=true
(via default parameter value) when retrieving the system DE key, and
likewise for read_or_create_volkey() and volume keys.
As a result, if the associated Keymaster key needs to be upgraded, the
upgraded key blob gets written to "keymaster_key_blob_upgraded", but it
doesn't replace the original "keymaster_key_blob", nor is the original
key deleted from Keymaster. This happens at every boot, eventually
resulting in the RPMB partition in Keymaster becoming full.
Only the metadata encryption key ever needs keepOld=true, since it's the
only key that isn't stored in /data, and the purpose of keepOld=true is
to allow a key that isn't stored in /data to be committed or rolled back
when a userdata checkpoint is committed or rolled back.
So, fix this bug by removing the default value of keepOld, and
specifying false everywhere except the metadata encryption key.
Note that when an affected device gets this fix, it will finally upgrade
its system DE key correctly. However, this fix doesn't free up space in
Keymaster that was consumed by this bug.
Test: On bramble:
- Flashed rvc-d1-dev build, with wiping userdata
- Flashed a newer build, without wiping userdata
- Log expectedly shows key upgrades:
$ adb logcat | grep 'Upgrading key'
D vold : Upgrading key: /metadata/vold/metadata_encryption/key
D vold : Upgrading key: /data/unencrypted/key
D vold : Upgrading key: /data/misc/vold/user_keys/de/0
D vold : Upgrading key: /data/misc/vold/user_keys/ce/0/current
- Rebooted
- Log unexpectedly shows the system DE key being upgraded again:
$ adb logcat | grep 'Upgrading key'
D vold : Upgrading key: /data/unencrypted/key
- "keymaster_key_blob_upgraded" unexpectedly still exists:
$ adb shell find /data /metadata -name keymaster_key_blob_upgraded
/data/unencrypted/key/keymaster_key_blob_upgraded
- Applied this fix and flashed, without wiping userdata
- Log shows system DE key being upgraded (expected because due to the
bug, the upgraded key didn't replace the original one before)
$ adb logcat | grep 'Upgrading key'
D vold : Upgrading key: /data/unencrypted/key
- "keymaster_key_blob_upgraded" expectedly no longer exists
$ adb shell find /data /metadata -name keymaster_key_blob_upgraded
- Rebooted
- Log expectedly doesn't show any more key upgrades
$ adb logcat | grep 'Upgrading key'
Bug: 171944521
Bug: 172019387
Change-Id: I42d3f5fbe32cb2ec229f4b614cfb271412a3ed29
We want various per-user directories to have their SELinux MLS level
set to restrict access from other users, as an improvement to user
isolation.
We extend vold_prepare_subdirs to implement this if a flag is
set. vold itself then sets the flag based on a new property,
ro.vold.level_from_user. This is to allow testing of further
incremental work to ensure system apps correctly handle the new
restriction on different devices rather than causing immediate
breakage. Eventually this will go away and the restriction will apply
everywhere.
Bug: 141677108
Test: Manual, with and without propery set.
Change-Id: I8e2207bd94b487bdcc09fd4d80b031027dfea1e3
The name "pre_gki_level" is causing some confusion because not all
devices launching with Android R are subject to the GKI requirement.
(See b/161563110#comment11.) E.g., devices that use a 4.14-based kernel
are exempt from GKI. However, the encryption requirements still apply.
Just use __ANDROID_API_Q__ directly instead.
No change in behavior.
Change-Id: Id02ae1140845ac1ae7cf78be4e57fe34da028abf
The emmc_optimized encryption flag is specifically designed for the
limitations of inline encryption hardware that follows the eMMC
standard. It isn't appropriate to use on other types of storage.
So, make vold enforce that it's not used on other types of storage.
Bug: 160639344
Test:
- Enabled emmc_optimized on Cuttlefish and verified it no longer boots
- Using a modified version of this change, verified that
IsEmmcStorage() works as expected on various devices including
Cuttlefish, Cuttlefish booted in GSI image mode, a device with eMMC
storage, and a device with UFS storage.
- Verified that VtsKernelEncryptionTest still passes
Change-Id: Ie27b80658db53b1a4207b3cbb4e309d05130812e
During userspace reboot /data might be unmounted & remounted, meaning
that CE keys stored in fs-level keyring will be lost. In order to be
able to restore them, when installing new key to fs-level keyring, it's
also added to session-level keyring with type "fscrypt-provisioning".
Then when init_user0 is called during userspace reboot, vold will try to
load CE keys from the session-level keyring back into fs-level keyring
for all the users that were unlocked before the reboot.
If for any user vold fails to install the key, init_user0 will fail and
fallback to hard reboot will be triggered.
Test: set a pin pattern
Test: adb shell setprop sys.init.userdata_remount.force_umount 1
Test: adb shell svc power reboot userspace
Test: atest CtsUserspaceRebootHostSideTestCases
Bug: 143970043
Change-Id: I37603dc136c7ededc7b0381e4d730cb0ffd912b4
Merged-In: I37603dc136c7ededc7b0381e4d730cb0ffd912b4
(cherry picked from commit 1ee35cf002)
Previously every directory on external storage had project ID quota
inheritance enabled; this means that if any new file/directory is
created under such a directory, it will inherit the project ID from the
parent. We use a default project ID of 1000 for generic directories, and
application-specific project IDs for app-specific directories.
MediaProvider is responsible for updating the quota type in the generic
directories, as it scans all files there. However, there is a problem
with this approach: if you move a file to a directory with project ID
inheritance set, and the project ID of that file differs from the
project ID of the dir, that results in an EXDEV error, and requires a
copy instead. For example, if /sdcard/DCIM/test.jpg has a project ID of
1003 (for images), and you try to move it to /sdcard/Pictures/test.jpg,
that would require a copy, because the project ID of /sdcard/Pictures is
1000.
While this is not a very common scenario, it's still better to avoid it.
Luckily we can - since MediaProvider anyway scans all files, it will set
the project ID on individual files correctly - there's no need to
inherit them.
We then only need to inherit quota in application-specific directories,
since in those directories the app can create files itself, and those
need to be tagged correctly.
This change enables that, by removing quota inheritance setting from the
top-level directory, and instead doing it for app-specific directories
instead.
Bug: 151078664
Test: atest StorageHostTest
atest com.android.tests.fused.host.FuseDaemonHostTest#testRenameAndReplaceFile
Change-Id: I38a057ec61cb627e39a3ff7ac58c7218dc251bdc
During userspace reboot /data might be unmounted & remounted, meaning
that CE keys stored in fs-level keyring will be lost. In order to be
able to restore them, when installing new key to fs-level keyring, it's
also added to session-level keyring with type "fscrypt-provisioning".
Then when init_user0 is called during userspace reboot, vold will try to
load CE keys from the session-level keyring back into fs-level keyring
for all the users that were unlocked before the reboot.
If for any user vold fails to install the key, init_user0 will fail and
fallback to hard reboot will be triggered.
Test: set a pin pattern
Test: adb shell setprop sys.init.userdata_remount.force_umount 1
Test: adb shell svc power reboot userspace
Test: atest CtsUserspaceRebootHostSideTestCases
Bug: 143970043
Change-Id: I37603dc136c7ededc7b0381e4d730cb0ffd912b4
During userspace reboot /data might be unmounted, which means that if
device supports filesystem keyring, DE keys will be lost and are needed
to be re-installed.
Test: adb shell setprop sys.init.userdata_remount.force_umount 1
Test: adb shell svc power reboot userspace
Test: atest CtsUserspaceRebootHostSideTestCases
Bug: 143970043
Change-Id: I153caa1d7c373b3c906a34f1184c681e52854a9d
Merged-In: I153caa1d7c373b3c906a34f1184c681e52854a9d
(cherry picked from commit 1eaea5a6a2)
During userspace reboot /data might be unmounted, which means that if
device supports filesystem keyring, DE keys will be lost and are needed
to be re-installed.
Test: adb shell setprop sys.init.userdata_remount.force_umount 1
Test: adb shell svc power reboot userspace
Test: atest CtsUserspaceRebootHostSideTestCases
Bug: 143970043
Change-Id: I153caa1d7c373b3c906a34f1184c681e52854a9d
Making HEH the default was always a mistake and a giant foot-gun.
Let's make life easier for people by making the default depend on
first_api_level, so it's automatically set up right for new devices
without breaking old ones. Also use v2 fscrypt keys instead of v1 post
Q.
Bug: 147107322
Test: Various Cuttlefish configurations
Change-Id: I5432bdfd6fec6ed34e7f9ab7cdd32cdeb2a03472
No need for KeyUtil to know how to make a KeyGeneration, it's cleaner
if each module handles it separately. Also, create a CryptoOptions
structure to track metadata encryption options, and simplify legacy
cipher/option handling.
Test: Treehugger
Bug: 147814592
Change-Id: I740063882914097329ff72348d0c0855c26c7aab
To prevent keys from being compromised if an attacker
acquires read access to kernel memory, some inline
encryption hardware supports protecting the keys in
hardware without software having access to or the
ability to set the plaintext keys. Instead, software
only sees "wrapped keys", which may differ on every boot.
'wrappedkey_v0' fileencryption flag is used to denote
that the device supports inline encryption hardware that
supports this feature. On such devices keymaster is used
to generate keys with STORAGE_KEY tag and export a
per-boot ephemerally wrapped storage key to install it in
the kernel.
The wrapped key framework in the linux kernel ensures the
wrapped key is provided to the inline encryption hardware
where it is unwrapped and the file contents key is derived
to encrypt contents without revealing the plaintext key in
the clear.
Test: FBE validation with Fscrypt v2 + inline crypt + wrapped
key changes kernel.
Bug: 147733587
Change-Id: I1f0de61b56534ec1df9baef075acb74bacd00758
When creating external storage directories like /data/media/0, make sure
we set the quota project ID correctly, and enable project ID
inheritance. This ensures that all directories/files under this will be
created with the correct project ID and inheritance as well.
Bug: 146419093
Test: lsattr -p on /data/media
Change-Id: I32bfced0d67eb8c1865897b085324f00c55926a0
Move most of it into cryptfs.cpp, and include cryptfs.h in fewer files.
Bug: 147814592
Test: Treehugger
Change-Id: Ia3592d73e7abc1f07a60538e0978a3033bdea7de
CE keys were cached in vold to support untrusted reset
by a device admin, this is now supported by Locksettingservice
using synthetic password. This change requires a secret to be
provided to retrieve the CE key and re-wrap without the secret
when user removes the credential.
Test: Set credential, remove credential, swipe to none
and vice-versa.
Bug: 26948053
Merged-In: I4cb1c035a472477e70c1ff5bf0b2c3fcfad495e5
Change-Id: I4cb1c035a472477e70c1ff5bf0b2c3fcfad495e5
Add support for setting v2 encryption policies when configured in the
fstab (for internal storage) or in system properties (for adoptable
storage), and for installing and evicting the keys for such policies.
v2 policies support the same encryption modes and flags as v1 policies,
but internally they use a more standard, secure, and flexible KDF. Due
to this, some future features will be supported by v2 policies only.
Bug: 140500999
Test: Configured a device to use v2 encryption policies (applied the
needed kernel patches and added
"fileencryption=aes-256-xts:aes-256-cts:v2" to fstab, and set the
corresponding system properties for adoptable storage). Wiped
userdata, booted device and checked logs to verify that v2
policies were being used.
Also enabled virtual SD card and formatted as adoptable storage;
verified it works and that v2 policies were being used on it.
Also created, started, and stopped a 2nd user and verified their
keys were evicted.
Also verified that the device comes up again after rebooting.
Also verified that a device using v1 encryption policies continues
to work, both with and without an updated kernel -- including
stopping a user so that their keys get evicted.
Change-Id: If64028d8580584b2c33c614cabd5d6b93657f608
When the kernel supports the new fscrypt key management ioctls, use them
instead of add_key() and keyctl_unlink().
This will be needed in order to support v2 encryption policies, since v2
encryption policies only support the new ioctls.
The new ioctls have other advantages too. For example,
FS_IOC_REMOVE_ENCRYPTION_KEY automatically evicts exactly the necessary
kernel objects, so the drop_caches sysctl is no longer needed. This
makes evicting keys faster and more reliable.
FS_IOC_REMOVE_ENCRYPTION_KEY also detects if any files are still open
and therefore couldn't be "locked", whereas this went undetected before.
Therefore, to start out this patch adds support for using the new ioctls
for v1 encryption policies, i.e. on existing devices.
(Originally based on a patch by Satya Tangirala <satyat@google.com>)
Bug: 140500828
Test: tested that a device using v1 policies continues to work, both
with and without an updated kernel. See
If64028d8580584b2c33c614cabd5d6b93657f608 for more details.
Also checked via the log that the filesystem-level keyring is in
fact used when supported.
Change-Id: I296ef78138578a3fd773797ac0cd46af1296b959
Bug: 140882488
Test: Booted twice, checked logs to ensure encryption
is different each time, adb created files in directory.
Change-Id: I74077bf8f727dab726c25f512ed7406a74cbe0bf
When Android boots after file_contexts has changed, the boot process
walks the entire /data partition, updating any changed SELinux labels as
appropriate. However, credential encrypted ("ce") directories are
deliberately excluded from this early boot directory walk. Files within
ce directories have their filenames encrypted, and as a result, cannot
match the file_contexts entries. Only after the user has unlocked their
device are the unencrypted filenames available and a restorecon
appropriate.
Ensure that we do a post-unlock restorecon on /data/vendor_ce, like we
do for /data/system_ce and /data/misc_ce. This ensures the labels on
files within these directories are correct after the device has been
unlocked.
Bug: 132349934
Test: See bug 132349934 comment #12 for test procedure
Change-Id: Ifcbef5fdfb236ec6dea418efa9d965db3a3b782f
When a user's CE key is removed, write "2" to /proc/sys/vm/drop_caches
rather than "3". This avoids unnecessarily evicting the pagecache of
in-use inodes. It's only necessary to evict the inodes of the relevant
encrypted files, and these are already sync'ed and no longer in-use.
For this mode "2" suffices, as this evicts "reclaimable slab objects",
including inodes; and evicting an inode implies evicting its pagecache.
This matches the recommendation I've made in the documentation for the
fscrypt kernel feature at
https://www.kernel.org/doc/html/latest/filesystems/fscrypt.html#online-attacks
Test: Sanity check that directories are still "locked" properly:
Unlock device with PIN. Then in adb shell: 'stop; start;
sleep 10; ls /data/data/' still shows filenames in ciphertext form.
Change-Id: I1bdf3c420ebf63e98cc314498211061ea36f2942
Use vold version of writeStringToFile which fsync files, and
manually fsync directories after initialize global DE
(cherry picked from commit a98464f688)
Bug: 71810347
Test: Build pass and reboot stress test.
Original boot failure symptom is NOT reproducible.
Change-Id: I1ca8f8cf0ccfd01075a9c33f79042e58d99aea26
Merged-In: I1ca8f8cf0ccfd01075a9c33f79042e58d99aea26
Now that lockUserKey() will be called for every user each time the
framework is started, it's a normal case that the key being removed
doesn't exist. So downgrade the log message about being unable to find
a key from ERROR to DEBUG.
Test: build and boot, message is now debug level.
Change-Id: I2d1dce8dd8825b02982eab1c902aa10fe5b54b84
Use vold version of writeStringToFile which fsync files, and
manually fsync directories after initialize global DE
Bug: 71810347
Test: Build pass and reboot stress test.
Original boot failure symptom is NOT reproducible.
Change-Id: I1ca8f8cf0ccfd01075a9c33f79042e58d99aea26
Device can't start up after the following steps:
1. set screen lock to PIN/Pattern/Password
2. set screen lock to Swipe/None
3. power down immediately after pressing "YES, REMOVE"
4. reboot
failed log:
Failed to read from /data/misc/vold/user_keys/ce/0/current/keymaster_key_blob
root cause:
flushing data failed because of power down
issue:
https://partnerissuetracker.corp.google.com/u/1/issues/119382750
We support file-based encryption on both ext4 and f2fs now, and the
kernel API is the same. So rename things appropriately in vold:
e4crypt => fscrypt
ext4enc => fscrypt
Ext4Crypt => FsCrypt
EXT4_* => FS_*
ext4_encryption_key => fscrypt_key
Additionally, the common functions shared by 'vold' and 'init' are now
in libfscrypt rather than ext4_utils. So update vold to link to
libfscrypt and include the renamed headers.
Note: there's a chance of 'fscrypt' being confused with the dm-crypt
based encryption code in vold which is called 'cryptfs'. However,
fscrypt is the name used in the kernel for ext4/f2fs/ubifs encryption,
and it's preferable to use the same name in userspace.
Test: built, booted device with f2fs encryption
Change-Id: I2a46a49f30d9c0b73d6f6fe09e4a4904d4138ff6
