Refactor fstrim code to be encapsulated in unique task object, and
give it option of benchmarking when finished. Trimming now includes
both storage from fstab and adopted private volumes. Cleaner timing
stats are logged for each unique volume.
Add wakelock during ongoing async move tasks. Push disk sysfs path
to framework so it can parse any SD card registers as desired.
Bug: 21831325
Change-Id: I76577685f5cae4929c251ad314ffdaeb5eb1c8bf
When formatting volumes, pass along fsType string which can be "auto"
to let the volume select the best choice. For now, private volumes
assume that MMC devices (like SD cards) are best off using f2fs when
both kernel support and tools are present, otherwise fall back to
ext4. Use blkid when mounting to pick the right set of tools.
Move filesystem utility methods into namespaces and place in separate
directory to be more organized.
Bug: 20275581
Change-Id: Id5f82d8672dda2e9f68c35b075f28232b0b55ed4
In order to prevent this bug from happening, we must allow vold cryptfs
commands to complete while a long running mount is underway.
While waiting for vold to be changed to a binder interface, we will simply
create two listeners, one for cryptfs and one for everything else.
Bug: 19197175
Change-Id: If74142aa81abd58f718a9de6c9e387f6ea442754
Now that we're offering to store private app data on adopted storage
devices, the performance of those devices is much more important to
overall user experience.
To help set user expectations, this change offers to execute a
real-world benchmark on a storage device, returning a metric that can
be used to compare internal and external storage. The benchmark is
generated from the strace-instrumented storage access patterns of
typical apps.
A typical device completes the benchmark in under two seconds on
internal storage, a UHS-3 SD card is even faster (!), but a very slow
Class 4 SD card takes about 30 seconds to complete, giving us a clear
signal.
The measured benchmark numbers are logged along with information
about the storage device, such as manufacturer, model, etc. Card
serial numbers are scrubbed from output.
Bug: 21172095
Change-Id: I9b2713dafdfdfcf5d97bf1bc21841f39409a7e54
When requested, kick off a thread that will migrate storage contents
between two locations. This is performed in several steps that
also interact with the framework:
1. Take old and new volumes offline during migration
2. Wipe new location clean (10% of progress)
3. Copy files from old to new (60% of progress)
4. Inform framework that move was successful so it can persist
5. Wipe old location clean (15% of progress)
Derives a hacky progress estimate by using a rough proxy of free
disk space changes while a cp/rm is taking place.
Add new internal path for direct access to volumes to bypass any
FUSE emulation overhead, and send it to framework. Remove mutex
around various exec calls since setexeccon() is already per-thread.
Bug: 19993667
Change-Id: Ibcb4f6fe0126d05b2365f316f53e71dc3e79a2b8
This adds support for private volumes which is just a filesystem
wrapped in a dm-crypt layer. For now we're using the exact same
configuration as internal encryption (aes-cbc-essiv:sha256), but we
don't store any key material on the removable media. Instead, we
store the key on internal storage, and use the GPT partition GUID
to identify which key should be used.
This means that private external storage is effectively as secure as
the internal storage of the device. That is, if the internal storage
is encrypted, then our external storage key is also encrypted.
When partitioning disks, we now support a "private" mode which has
a PrivateVolume partition, and a currently unused 16MB metadata
partition reserved for future use. It also supports a "mixed" mode
which creates both a PublicVolume and PrivateVolume on the same
disk. Mixed mode is currently experimental.
For now, just add ext4 support to PrivateVolume; we'll look at f2fs
in a future change. Add VolumeBase lifecycle for setting up crypto
mappings, and extract blkid logic into shared method. Sprinkle some
more "static" around the cryptfs code to improve invariants.
Bug: 19993667
Change-Id: Ibd1df6250735b706959a1eb9d9f7219ea85912a0
This is the first in a series of changes that are designed to
introduce better support for dynamic block devices.
It starts by defining a new Volume object which represents a storage
endpoint that knows how to mount, unmount, and format itself. This
could be a filesystem directly on a partition, or it could be an
emulated FUSE filesystem, an ASEC, or an OBB.
These new volumes can be "stacked" so that unmounting a volume will
also unmount any volumes stacked above it. Volumes that provide
shared storage can also be asked to present themselves (through bind
mounts) into user-specific mount areas.
This change also adds a Disk class which is created based on block
kernel netlink events. Instead of waiting for partition events from
the kernel, it uses gptfdisk to read partition details and creates
the relevant Volume objects.
Change-Id: I0e8bc1f8f9dcb24405f5e795c0658998e22ae2f7
The libcrypto and libssl modules (and their respective static and host
versions) use LOCAL_EXPORT_C_INCLUDE_DIRS thus just including the module
is sufficient.
Additionally, cryptfs.h was including an OpenSSL header just to get the
length of a SHA-256 hash. Rather than force all users of this header to
also depend on libcrypto, it's easier just to define that value in the
header file.
Change-Id: I3e3e0db906a212e1093944b298e4a8ff2e2fb07d
Since the dm_ioctl struct was being allocated on the stack as a large
character array, it was getting character alignment rather than the
proper alignment for the struct. GCC had been getting away with this
so far, but it's undefined behavior that clang managed to expose.
Bug: 18736778
Change-Id: Ied275dfad7fcc41d712b2d02c8a185f499221f57
-Wno-missing-field-initializers is used as well, but that is an
overzealous warning from initializing structs with {0} and not a
real warning.
bug 18736778 and 16868177
Change-Id: Iffde89cd7200d9a11193e1614f1819f9fcace30a
It looks like clang might have a miscompile that is causing SIGBUS in
`ioctl_init` when the device is encrypted. Move back to GCC until we
can sort this out.
Bug: 18736778
Change-Id: I21ae3b9d7d9ebff8679ecc1a828b7c59f27d0903
Currently Android provides disk encryption support using dm-crypt
which is based on bios. dm-crypt uses 512 bytes packet size for
crypto operations. While 512 bytes size packet is ok for SW based
disk encryption, it is inefficient for HW based crypto engines.
dm-req-crypt is similar to dm-crypt except it uses block requests
rathe bios for crypto operations. block requests when unpacked
carries data upto 512KB. Hence, HW based crypto engine can be used
more efficiently.
Also move create disk encryption key before framework start as
HW based disk encryption creates key in secure side. Key creation
can take sometime to create the key securely. If framework is
started before creating the key, it is possible that framework
requests service from secure side. Secure side can serve mostly one
request at a time. Secure side may reject framework request if key
creation request is still going on. This may cause problem in the
system
b/17475056 Enable hardware crypto for userdata encryption
Change-Id: I5480ab72a37c02532218a18faaba598a824589fd
Signed-off-by: Iliyan Malchev <malchev@google.com>
ASAN_ALL uses ASAN for anything built with clang. Since some of vold's
dependencies use clang, they will have unresolved ASAN symbols unless
vold is also built with clang. There's no harm in just moving this
project to clang.
Change-Id: Ia6f412beb7bf092121bff2a5a980531636adcdb9
Sami Tolvanen
Tom Cherry
Jeff Sharkey
Paul Lawrence
Kenny Root
Paul Crowley
Mohamad Ayyash
Dan Albert
Adam Langley
Tim Murray
Iliyan Malchev
Ajay Dudani
Daniel Rosenberg
Jim Miller