OSDN Git Service

android-x86/system-vold.git
8 years agovold: Remove creation of new keymaster key for password update
Dinesh K Garg [Fri, 2 Oct 2015 20:38:18 +0000 (13:38 -0700)]
vold: Remove creation of new keymaster key for password update

To update user password, VOLD creates a new keymaster key which
would be used in the process to rewrap the FDE key. These already
exists a keymaster blob which contains keymaster key created for
old password. This old keymaster blob is secure and there is no
need to create new keymaster key which can take sometime long time
as well. Hence, removing the need to create a new keymaster key which
does not enhance security but sometimes take longer time to update
user password.

Change-Id: I6ce0f71353d54cb32a11d44731b30fa41df235c8

8 years agovold: Fix return value from get_keymaster_hw_fde_passwd()
Dinesh K Garg [Mon, 28 Sep 2015 17:56:00 +0000 (10:56 -0700)]
vold: Fix return value from get_keymaster_hw_fde_passwd()

get_keymaster_hw_fde_passwd() is returing wrong value when HW FDE keys
are not tied to keymaster. Fixed the return values.

Change-Id: Icc5850b0f3154ab8e685f69f02d2a9ae17048528

8 years agoAdded cypto_complete states to reflect MDTP state.
Amit Blay [Mon, 31 Aug 2015 07:16:04 +0000 (10:16 +0300)]
Added cypto_complete states to reflect MDTP state.

Also, fixed bug in the wipe option in cryptfs_enable_internal.

Change-Id: I5b441df6fd55b9c1d40797b25c4a5d6b3bfddcc8

8 years agovold: Fix HW FDE OTA support on SW FDE encrypted device
Dinesh K Garg [Fri, 18 Sep 2015 16:55:53 +0000 (09:55 -0700)]
vold: Fix HW FDE OTA support on SW FDE encrypted device

If an OTA is delivered to a device encrypted with SW FDE, an OTA
of HW FDE should not force user to switch to HW FDE and hence wipe
of all userdata. Instead, it should continue using SW FDE until user
itself perform factory reset. After factory reset, HW FDE would be
used to encrypt the data. This support somehow was broken. This
change fixes the support.

Change-Id: If31ccbd377e5a161a90b2e4a2c53dcc52712b49c

8 years agovold: Tie HW FDE keys with Root of Trust(ROT)
Dinesh K Garg [Thu, 13 Aug 2015 18:38:16 +0000 (11:38 -0700)]
vold: Tie HW FDE keys with Root of Trust(ROT)

HW FDE keys should be tied with device's ROT. Adding required changes
to address the new requirement.

Change-Id: Id2718b48140b97868bcaef90ab9357e4e0327da6

8 years agovold: Port HW FDE features
Dinesh K Garg [Thu, 6 Aug 2015 23:31:16 +0000 (16:31 -0700)]
vold: Port HW FDE features

Following changes for HW FDE are missing which needs to be ported:
- Restart Android framework after HW FDE key has been created
- Add support of Inline Cryto Engine
- Use new HW FDE apis to update password

Change-Id: I49f16867d45a8c10fe7239f600170fd636ac3eda

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 13 Aug 2015 08:08:30 +0000 (01:08 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoRequest specific tags from blkid.
Jeff Sharkey [Wed, 12 Aug 2015 23:04:35 +0000 (16:04 -0700)]
Request specific tags from blkid.

Otherwise blkid can emit tags like SEC_TYPE which mess with the value
extraction code.

Bug: 23069906
Change-Id: Id2a588ff43a538747d1e44cd8218c96ebd0192c2

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 8 Aug 2015 08:09:57 +0000 (01:09 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoProtect runtime storage mount points.
Jeff Sharkey [Thu, 6 Aug 2015 18:40:00 +0000 (11:40 -0700)]
Protect runtime storage mount points.

We have a bunch of magic that mounts the correct view of storage
access based on the runtime permissions of an app, but we forgot to
protect the real underlying data sources; oops.

This series of changes just bumps the directory heirarchy one level
to give us /mnt/runtime which we can mask off as 0700 to prevent
people from jumping to the exposed internals.

Also add CTS tests to verify that we're protecting access to
internal mount points like this.

Bug: 22964288
Change-Id: I83f09f0423f4993e766273c50389dd29b1c50589

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 1 Aug 2015 08:11:11 +0000 (01:11 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoReturn useful path when not visible.
Jeff Sharkey [Thu, 30 Jul 2015 23:54:23 +0000 (16:54 -0700)]
Return useful path when not visible.

This allows apps like ExternalStorageProvider to still read/write
files on transient storage devices which aren't mounted as visible.

Bug: 22545248
Change-Id: Idacb15f2233245a8e1861d9be977535a82b218ec

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 30 Jul 2015 08:06:06 +0000 (01:06 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoUse random data for benchmark instead of zeros.
Jeff Sharkey [Wed, 29 Jul 2015 16:14:21 +0000 (09:14 -0700)]
Use random data for benchmark instead of zeros.

If we always write zeros, we're leaving a giant pile of known
plaintext at an almost deterministic location on newly formatted
volumes.  To avoid this, repeat a 64K chunk of random data.

Bug: 22816936
Change-Id: Iedc067a519bd676a93b9d74ea4f9f77c84c8461c

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Wed, 29 Jul 2015 08:10:14 +0000 (01:10 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoGive secondary users read-only physical cards.
Jeff Sharkey [Tue, 28 Jul 2015 17:57:29 +0000 (10:57 -0700)]
Give secondary users read-only physical cards.

Long ago, we mounted secondary physical cards as readable by all
users on the device, which enabled the use-case of loading media on
a card and viewing it from all users.

More recently, we started giving write access to these secondary
physical cards, but this created a one-directional channel for
communication across user boundaries; something that CDD disallows.

This change is designed to give us the best of both worlds: the
package-specific directories are writable for the user that mounted
the card, but access to those "Android" directories are blocked for
all other users.  Other users remain able to read content elsewhere
on the card.

Bug: 22787184
Change-Id: I75dbd339f11402ae774c7e4b8f2b15ee216270e8

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 23 Jul 2015 08:07:03 +0000 (01:07 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoCreate user directory on emulated storage.
Jeff Sharkey [Tue, 21 Jul 2015 21:22:01 +0000 (14:22 -0700)]
Create user directory on emulated storage.

When mounting a primary external storage device that is multi-user
aware, ensure that the user-specific directory actually exists before
moving forward.

Bug: 22472026
Change-Id: I33c8eed261a9c0d5acedd5be6133ed9990679d08

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 9 Jul 2015 08:05:04 +0000 (01:05 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoMount emulated volumes as consistent location.
Jeff Sharkey [Tue, 7 Jul 2015 21:37:03 +0000 (14:37 -0700)]
Mount emulated volumes as consistent location.

We really only support a single emulated volume on the device at a
time, either on internal storage, or moved to a private volume.  To
avoid kicking off a giant rescan of all media when moved, keep all
the paths the same when mounted as primary.

Also ensure we have /data/media/0 ready on private volumes.

Bug: 20275423
Change-Id: I0c102f430b865ca7536772b1fae56d8c9660a97a

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 2 Jul 2015 08:02:41 +0000 (01:02 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoTrim both internal and adopted private storage.
Jeff Sharkey [Fri, 26 Jun 2015 21:02:09 +0000 (14:02 -0700)]
Trim both internal and adopted private storage.

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

8 years agoNull-terminate readlink() result, full remount.
Jeff Sharkey [Tue, 30 Jun 2015 22:54:17 +0000 (15:54 -0700)]
Null-terminate readlink() result, full remount.

In order to compare results from readlink() calls, we need to null
terminate the read value, otherwise we can end up doing an infinitely
recursive remount in the root namespace.

When remounting inside a namespace, unmount all existing mounts before
mounting the new storage into place.  This also means we need to mount
the user-specific symlinks back into place.

Skip spinning up the FUSE daemon when not visible, otherwise we get
stuck waiting for a daemon that never shows up.

Bug: 2219251822204412
Change-Id: Icc7db822354ab7ffc47c39cd0611f65edecc32e5

8 years agoNull-terminate readlink() result, full remount.
Jeff Sharkey [Tue, 30 Jun 2015 22:54:17 +0000 (15:54 -0700)]
Null-terminate readlink() result, full remount.

In order to compare results from readlink() calls, we need to null
terminate the read value, otherwise we can end up doing an infinitely
recursive remount in the root namespace.

When remounting inside a namespace, unmount all existing mounts before
mounting the new storage into place.  This also means we need to mount
the user-specific symlinks back into place.

Skip spinning up the FUSE daemon when not visible, otherwise we get
stuck waiting for a daemon that never shows up.

Bug: 2219251822204412
Change-Id: Icc7db822354ab7ffc47c39cd0611f65edecc32e5

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 27 Jun 2015 08:06:10 +0000 (01:06 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoLast ditch check for partition on raw disk.
Jeff Sharkey [Fri, 26 Jun 2015 18:16:14 +0000 (11:16 -0700)]
Last ditch check for partition on raw disk.

Some storage devices can be formatted as bare partitions, without an
MBR or GPT header.  If we found no partitions, try poking at the raw
disk, and treat it as a public volume if we found a valid filesystem.

Bug: 20503551
Change-Id: I80304e1ad865435321c341b667c0daf2daf4d56c

8 years agoLet's reinvent storage, yet again!
Jeff Sharkey [Wed, 24 Jun 2015 18:49:24 +0000 (11:49 -0700)]
Let's reinvent storage, yet again!

Now that we're treating storage as a runtime permission, we need to
grant read/write access without killing the app.  This is really
tricky, since we had been using GIDs for access control, and they're
set in stone once Zygote drops privileges.

The only thing left that can change dynamically is the filesystem
itself, so let's do that.  This means changing the FUSE daemon to
present itself as three different views:

/mnt/runtime_default/foo - view for apps with no access
/mnt/runtime_read/foo - view for apps with read access
/mnt/runtime_write/foo - view for apps with write access

There is still a single location for all the backing files, and
filesystem permissions are derived the same way for each view, but
the file modes are masked off differently for each mountpoint.

During Zygote fork, it wires up the appropriate storage access into
an isolated mount namespace based on the current app permissions.  When
the app is granted permissions dynamically at runtime, the system
asks vold to jump into the existing mount namespace and bind mount
the newly granted access model into place.

Bug: 21858077
Change-Id: Iade538e4bc7af979fe20095f74416e8a0f165a4a

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 25 Jun 2015 08:03:56 +0000 (01:03 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoMerge "Make sure volume is vfat before fsck'ing." into mnc-dev
Makoto Onuki [Wed, 24 Jun 2015 20:33:55 +0000 (20:33 +0000)]
Merge "Make sure volume is vfat before fsck'ing." into mnc-dev

8 years agoMake sure volume is vfat before fsck'ing.
Makoto Onuki [Wed, 24 Jun 2015 20:30:45 +0000 (13:30 -0700)]
Make sure volume is vfat before fsck'ing.

Bug 21948137

Change-Id: I6843423fd8809d9e2f352059a810aa17dd83b3e3

8 years agoAdd purpose to vold-generated keymaster1 keys.
Shawn Willden [Wed, 24 Jun 2015 14:21:54 +0000 (07:21 -0700)]
Add purpose to vold-generated keymaster1 keys.

Also remove the app ID and additional padding and digest options.

Bug: 22009890
Change-Id: Ibff9bbd0e0c11d651d11fac85d4ac907588f1cd2

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Tue, 23 Jun 2015 08:09:10 +0000 (01:09 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoAdd method to forget private partition keys.
Jeff Sharkey [Thu, 18 Jun 2015 21:25:08 +0000 (14:25 -0700)]
Add method to forget private partition keys.

Report both the disk and the partition GUID for private volumes to
userspace, and offer to forget the encryption key for a given
partition GUID.

Bug: 21782268
Change-Id: Ie77a3a58e47bf3563cdb3e4b0edfab1de4d0e6b4

8 years agoUse correct error code for rate limiting.
Shawn Willden [Thu, 18 Jun 2015 18:26:59 +0000 (12:26 -0600)]
Use correct error code for rate limiting.

Note that this CL depends on cl 712195, which must be submitted first.

Bug: 21607106
Change-Id: Iafc42d1c8a1145a31ea252b33b404044f92ec62b

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 18 Jun 2015 08:03:59 +0000 (01:03 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoAdd keymaster1 support to vold.
Shawn Willden [Wed, 3 Jun 2015 15:40:45 +0000 (09:40 -0600)]
Add keymaster1 support to vold.

Bug: 21607106
Change-Id: I498141b90888d4f0652912413b04519f61886935

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Tue, 16 Jun 2015 08:06:27 +0000 (01:06 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Mon, 15 Jun 2015 18:58:52 +0000 (11:58 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 13 Jun 2015 08:08:21 +0000 (01:08 -0700)]
merge in mnc-release history after reset to mnc-dev

8 years agoImprove boot time by 0.1s by reducing a polling sleep interval
Paul Lawrence [Thu, 11 Jun 2015 18:15:29 +0000 (11:15 -0700)]
Improve boot time by 0.1s by reducing a polling sleep interval

Bug: 21516860
Change-Id: I9e28f4d9cc20ec2a7d9e325c02ef85f0ad9b3d60

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 11 Jun 2015 08:03:52 +0000 (01:03 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoMerge "Start tracking added users with serial numbers." into mnc-dev
Jeff Sharkey [Wed, 10 Jun 2015 16:44:50 +0000 (16:44 +0000)]
Merge "Start tracking added users with serial numbers." into mnc-dev

9 years agoStart tracking added users with serial numbers.
Jeff Sharkey [Wed, 10 Jun 2015 16:42:01 +0000 (09:42 -0700)]
Start tracking added users with serial numbers.

vold will eventually use the serial numbers to clean up stale user
directories when mounting private storage devices.

Bug: 20275572
Change-Id: Ia29cb5da23e969f3087bb5caa5dc8f4e88f07613

9 years agoRemove hex encoding and password adjusting now that patterns are '1' based
Paul Lawrence [Tue, 9 Jun 2015 20:37:44 +0000 (13:37 -0700)]
Remove hex encoding and password adjusting now that patterns are '1' based

Bug: 21606650
Change-Id: I3486ad394d563135c5171a1d4785f7a27eeea3ae

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Wed, 10 Jun 2015 08:05:40 +0000 (01:05 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoRun restorecon over mounted private volumes.
Jeff Sharkey [Tue, 9 Jun 2015 17:59:17 +0000 (10:59 -0700)]
Run restorecon over mounted private volumes.

This ensures that we have consistent SELinux policy in place before
going any further, and it mirrors the way we restorecon /data when
first mounted.

Bug: 21121357
Change-Id: I2a7e3584ade655fe1fae8916cf54f9eae3a0f99d

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Tue, 9 Jun 2015 08:05:46 +0000 (01:05 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoAdd f2fs support for private volumes.
Jeff Sharkey [Fri, 22 May 2015 05:35:42 +0000 (22:35 -0700)]
Add f2fs support for private volumes.

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

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 6 Jun 2015 08:06:16 +0000 (01:06 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoMove crypt commands to a different listener in vold
Paul Lawrence [Wed, 3 Jun 2015 21:19:51 +0000 (14:19 -0700)]
Move crypt commands to a different listener in vold

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

9 years agoImprove boot time
Paul Lawrence [Thu, 4 Jun 2015 17:49:25 +0000 (10:49 -0700)]
Improve boot time

Bug: 21516860
Change-Id: I3153c7f9a414a8eeadc0118f0a642ad8e96c81ce

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 30 May 2015 08:08:28 +0000 (01:08 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoMerge "Don't use TEMP_FAILURE_RETRY on close in vold." into mnc-dev
Elliott Hughes [Fri, 29 May 2015 22:04:26 +0000 (22:04 +0000)]
Merge "Don't use TEMP_FAILURE_RETRY on close in vold." into mnc-dev

9 years agoDO NOT MERGE Delete password as per block encryption
Paul Lawrence [Wed, 6 May 2015 20:53:43 +0000 (13:53 -0700)]
DO NOT MERGE Delete password as per block encryption

(cherry-picked from commit 00f4aade5c172534c16070540d1c6c26d0a78c84)

Bug: 18151196
Change-Id: Iee0f932c61ff4a309dc2861725b24bf976adb4c7

9 years agoDO NOT MERGE Fix problem that reading/writing crypto footers wasn't identity
Paul Lawrence [Tue, 5 May 2015 22:58:27 +0000 (15:58 -0700)]
DO NOT MERGE Fix problem that reading/writing crypto footers wasn't identity

(cherry-picked from commit 75c922f49b593f3203ee59c33d00ccb5b3b66ca8)

Bug: 18151196
Change-Id: Ideef6bcdbccf068a64ed3e042be50c4837a373f8

9 years agoDO NOT MERGE Retry unmounts in ext4 encryption
Paul Lawrence [Tue, 5 May 2015 21:28:25 +0000 (14:28 -0700)]
DO NOT MERGE Retry unmounts in ext4 encryption

(cherry-picked from commit 29b54aab8ee2d08e2129832364f9b719dd17fa4e)

Bug: 18151196
Change-Id: I52ca23b2ce3adcff44bd003d4a12243a0bd6ac34

9 years agoDO NOT MERGE Use default key permissions for ext4enc
Paul Lawrence [Tue, 5 May 2015 18:09:07 +0000 (11:09 -0700)]
DO NOT MERGE Use default key permissions for ext4enc

(cherry-picked from commit 1190a26f6d7dd34bf10ffc7b367b374d5e591146)

As per discussion default permissions are the correct ones.
Note that since we use logon keys, they cannot be read outside
the kernel.

Note also that we limit who can read/write keys in selinux policy.

Bug: 18151196
Change-Id: Icc916f430a70eff22e6b74c20ec361c8f3789c1c

9 years agoDO NOT MERGE Simplify password checking logic
Paul Lawrence [Mon, 4 May 2015 22:48:24 +0000 (15:48 -0700)]
DO NOT MERGE Simplify password checking logic

(cherry-picked from commit aaccfac3442ab48f5f94603c07427cc1bb00e12d)

Bug: 18151196
Change-Id: I07ffde534dee7d1032149cfcbaa1a61c5246d759

9 years agoDO NOT MERGE Enable properties in ext4enc
Paul Lawrence [Wed, 15 Apr 2015 21:12:00 +0000 (14:12 -0700)]
DO NOT MERGE Enable properties in ext4enc

(cherry-picked from 4e7274551c93e1c064648409f52ca430da647050)

Enables OwnerInfo and pattern suppression

Bug: 18151196

Change-Id: I46144e16cb00319deeb5492ab82c67f5dd43d6d3

9 years agoDO NOT MERGE Check password is correct by checking hash
Paul Lawrence [Tue, 14 Apr 2015 22:26:29 +0000 (15:26 -0700)]
DO NOT MERGE Check password is correct by checking hash

(cherry-picked from commit 3ca21e227a2e1ed01138a29f450917290a9d1e6e)

Handle failures gracefully

Change-Id: Ifb6da8c11a86c50fb11964c18cc1be1326461f78

9 years agoDO NOT MERGE New ext4enc kernel switching from xattrs to ioctl
Paul Lawrence [Fri, 10 Apr 2015 14:48:51 +0000 (07:48 -0700)]
DO NOT MERGE New ext4enc kernel switching from xattrs to ioctl

(cherrypicked from commit 5e7f0042318156ef39c1e4641613f3191ea85b4a)

This is one of three changes to enable this functionality:
  https://android-review.googlesource.com/#/c/146259/
  https://android-review.googlesource.com/#/c/146264/
  https://android-review.googlesource.com/#/c/146265/

Bug: 18151196

Change-Id: Iba5146b8be1e15050ae901e08b3aaa26d96dcf7e

9 years agoDO NOT MERGE Securely encrypt the master key
Paul Lawrence [Tue, 28 Apr 2015 22:14:15 +0000 (22:14 +0000)]
DO NOT MERGE Securely encrypt the master key

(cherry-picked from commit 707fd6c7cccc31c0ab0ec1a6ac8b6077c632fc35)

Move all key management into vold
Reuse vold's existing key management through the crypto footer
to manage the device wide keys.

Use ro.crypto.type flag to determine crypto type, which prevents
any issues when running in block encrypted mode, as well as speeding
up boot in block or no encryption.

This is one of four changes to enable this functionality:
  https://android-review.googlesource.com/#/c/148586/
  https://android-review.googlesource.com/#/c/148604/
  https://android-review.googlesource.com/#/c/148606/
  https://android-review.googlesource.com/#/c/148607/

Bug: 18151196

Change-Id: I3c68691717a61b5e1df76423ca0c02baff0dab98

9 years agoDon't use TEMP_FAILURE_RETRY on close in vold.
Elliott Hughes [Fri, 29 May 2015 05:02:14 +0000 (22:02 -0700)]
Don't use TEMP_FAILURE_RETRY on close in vold.

Bug: http://b/20501816
Change-Id: Ia0a3899947582b8dd0e8e185c203d1ad7dad572e

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Wed, 20 May 2015 08:05:48 +0000 (01:05 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoOffer debug flag to force default primary.
Jeff Sharkey [Tue, 19 May 2015 22:54:15 +0000 (15:54 -0700)]
Offer debug flag to force default primary.

Useful for transforming a secondary SD card slot into a primary SD
card slot without doing a completely different lunch target.

$ adb shell stop
$ adb shell setprop vold.debug.default_primary 1
$ adb shell setprop ro.vold.primary_physical 1
$ adb shell setprop ctl.restart vold
$ adb shell start

Bug: 21017105
Change-Id: I45aac7ab3bb4c5ac3e0140572844883823fc3e4f

9 years agoOffer debug flag to force default primary.
Jeff Sharkey [Tue, 19 May 2015 22:54:15 +0000 (15:54 -0700)]
Offer debug flag to force default primary.

Useful for transforming a secondary SD card slot into a primary SD
card slot without doing a completely different lunch target.

$ adb shell stop
$ adb shell setprop vold.debug.default_primary 1
$ adb shell setprop ro.vold.primary_physical 1
$ adb shell setprop ctl.restart vold
$ adb shell start

Bug: 21017105
Change-Id: I45aac7ab3bb4c5ac3e0140572844883823fc3e4f

9 years agoFully sync filesystem before time measurements.
Jeff Sharkey [Tue, 19 May 2015 18:20:48 +0000 (11:20 -0700)]
Fully sync filesystem before time measurements.

Also drop unneeded SCR register logging.

Bug: 21172095
Change-Id: I17c5aae2edc0eb7144ab38701f90a1fd54e119da

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Sat, 16 May 2015 08:04:46 +0000 (01:04 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoMerge "Don't use TEMP_FAILURE_RETRY on close in vold." into mnc-dev
Elliott Hughes [Sat, 16 May 2015 02:42:30 +0000 (02:42 +0000)]
Merge "Don't use TEMP_FAILURE_RETRY on close in vold." into mnc-dev

9 years agoDon't use TEMP_FAILURE_RETRY on close in vold.
Elliott Hughes [Sat, 16 May 2015 01:34:24 +0000 (18:34 -0700)]
Don't use TEMP_FAILURE_RETRY on close in vold.

Bug: http://b/20501816
Change-Id: Ieecce9304539c250ed1728252b8c2c09d29afd7f

9 years agoInitial pass at storage benchmarks.
Jeff Sharkey [Fri, 15 May 2015 03:33:55 +0000 (20:33 -0700)]
Initial pass at storage benchmarks.

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

9 years agomerge in mnc-release history after reset to mnc-dev
The Android Automerger [Thu, 14 May 2015 08:05:50 +0000 (01:05 -0700)]
merge in mnc-release history after reset to mnc-dev

9 years agoCommunicate if we have adoptable disks.
Jeff Sharkey [Wed, 13 May 2015 20:53:07 +0000 (13:53 -0700)]
Communicate if we have adoptable disks.

Bug: 19993667
Change-Id: I3c3dddab157a11038f7160d24d53e728bde2194f

9 years agoReport internal path for public volumes.
Jeff Sharkey [Wed, 13 May 2015 19:36:48 +0000 (12:36 -0700)]
Report internal path for public volumes.

Used by framework to bypass FUSE daemon overhead in some cases.

Bug: 21017105
Change-Id: I2f3ae70607417fc4cd2c970cb17cf3afa2f42613

9 years agoReport internal path for public volumes.
Jeff Sharkey [Wed, 13 May 2015 19:36:48 +0000 (12:36 -0700)]
Report internal path for public volumes.

Used by framework to bypass FUSE daemon overhead in some cases.

Bug: 21017105
Change-Id: I2f3ae70607417fc4cd2c970cb17cf3afa2f42613

9 years agoMerge "Fix signed issue with hex conversion." into mnc-dev
Jeff Sharkey [Thu, 30 Apr 2015 17:07:15 +0000 (17:07 +0000)]
Merge "Fix signed issue with hex conversion." into mnc-dev

9 years agoUpdate primary symlinks after mounting.
Jeff Sharkey [Wed, 29 Apr 2015 22:22:23 +0000 (15:22 -0700)]
Update primary symlinks after mounting.

Since otherwise we might have a stale path.

Bug: 19993667
Change-Id: I099e3dc0c5aa9ab6820cded4f2ae7fd6bf18ea40

9 years agoFix signed issue with hex conversion.
Jeff Sharkey [Wed, 29 Apr 2015 22:57:48 +0000 (15:57 -0700)]
Fix signed issue with hex conversion.

Change-Id: Ia11c689c373f41b2a63bc84151eb16f7c7b9d155

9 years agoProperty to enable debugging.
Jeff Sharkey [Wed, 29 Apr 2015 15:57:18 +0000 (08:57 -0700)]
Property to enable debugging.

Bug: 19993667
Change-Id: Ie226e06edcb426caca2dbc97d2d5d60ae948c85e

9 years agoScan failure still means we finished scan.
Jeff Sharkey [Wed, 29 Apr 2015 04:17:43 +0000 (21:17 -0700)]
Scan failure still means we finished scan.

Bug: 19993667
Change-Id: If6a79391e34b652b2ba4e6df40a8e712f7a2e172

9 years agoam 85c7c08e: am d5aa67cc: Merge "Wait for completion of device mapping in mountObb"
Elliott Hughes [Sat, 25 Apr 2015 17:24:04 +0000 (17:24 +0000)]
am 85c7c08e: am d5aa67cc: Merge "Wait for completion of device mapping in mountObb"

* commit '85c7c08e2511509a306c3a5b202ceda922f5f2b5':
  Wait for completion of device mapping in mountObb

9 years agoam d5aa67cc: Merge "Wait for completion of device mapping in mountObb"
Elliott Hughes [Sat, 25 Apr 2015 17:02:47 +0000 (17:02 +0000)]
am d5aa67cc: Merge "Wait for completion of device mapping in mountObb"

* commit 'd5aa67cc53bd10ba851179d634a7ed649958e5ee':
  Wait for completion of device mapping in mountObb

9 years agoMerge "Wait for completion of device mapping in mountObb"
Elliott Hughes [Sat, 25 Apr 2015 16:42:37 +0000 (16:42 +0000)]
Merge "Wait for completion of device mapping in mountObb"

9 years agoWait for completion of device mapping in mountObb
yoshiyuki hama [Wed, 28 Jan 2015 07:37:23 +0000 (16:37 +0900)]
Wait for completion of device mapping in mountObb

The VolumeManager::mountObb() creates a mapping between
a loopback device and a dm device. However the device-mapper
carries it out asynchronously, so there is a possibility that
Vold accesses to the dm device which is being built. Added
waiting for completion of the mapping in that function, like
mountAsec().

To verify install FrameworksCoreTests.apk and do:

  adb shell am instrument -r -w -e class android.os.storage.\
  StorageManagerIntegrationTest#testMountTwoEncryptedObb \
  com.android.frameworks.coretests/android.test.\
  InstrumentationTestRunner

Change-Id: If42f4b7494bb2f8a8b72d106ad84b3e3bf91fd9b

9 years agoMigrate primary external storage.
Jeff Sharkey [Fri, 24 Apr 2015 23:00:03 +0000 (16:00 -0700)]
Migrate primary external storage.

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

9 years agoWider volume mutation lock, move force adoptable.
Jeff Sharkey [Tue, 21 Apr 2015 19:14:17 +0000 (12:14 -0700)]
Wider volume mutation lock, move force adoptable.

We eventually should move back to per-disk locks, but use a giant
lock to keep development rolling forward.  Also move force adoptable
flag to framework since, since encrypted devices don't have persisted
properties loaded early during boot.

Bug: 19993667
Change-Id: Ifa3016ef41b038f8f71fc30bc81596cfd21dcd2a

9 years agoHandle *ALL* the SCSI disks!
Jeff Sharkey [Sun, 19 Apr 2015 22:55:42 +0000 (15:55 -0700)]
Handle *ALL* the SCSI disks!

Bug: 19993667
Change-Id: I47099c262686127f82dd376570d49cf0ad119842

9 years agoNotify when disk scan finished.
Jeff Sharkey [Sun, 19 Apr 2015 21:57:55 +0000 (14:57 -0700)]
Notify when disk scan finished.

Bug: 19993667
Change-Id: I20dc183a35c11b5b4ba6139e690929b9c279b830

9 years agoSwitch logging to system log, persistable prop.
Jeff Sharkey [Sat, 18 Apr 2015 23:15:10 +0000 (16:15 -0700)]
Switch logging to system log, persistable prop.

Use system log for more durable log history.  Also accept persistable
version of flag to force adoptability.

Bug: 19993667
Change-Id: Icdf36799e26038e9a8e2bdaca276d97b22ba3941

9 years agoVolumes know parent disks; unsupported disks.
Jeff Sharkey [Sat, 18 Apr 2015 00:35:20 +0000 (17:35 -0700)]
Volumes know parent disks; unsupported disks.

This is cleaner and more direct than the reverse of having the disk
publish child volume membership.  Rename state constants to match
public API.  Add state representing bad removal.  Make it clear that
volume flags are related to mounting.

Send new unsupported disk event when we finish scanning an entire
disk and have no meaningful volumes.

Bug: 19993667
Change-Id: I08a91452ff561171a484d1da5745293ec893aec0

9 years agoLock while partitioning.
Jeff Sharkey [Wed, 15 Apr 2015 06:14:23 +0000 (23:14 -0700)]
Lock while partitioning.

Otherwise we get really excited and trip over ourselves while
partitions are still being created.

Bug: 19993667
Change-Id: I034e56b3063a71d73f9311a945c05ea2ae255f7d

9 years agoDon't be picky about zap status.
Jeff Sharkey [Wed, 15 Apr 2015 05:22:34 +0000 (22:22 -0700)]
Don't be picky about zap status.

Bug: 19993667
Change-Id: Id125d9252bf7130516c0ec619d5f067fea24d560

9 years agoFix destroy ordering bug; add debug flag.
Jeff Sharkey [Tue, 14 Apr 2015 23:45:18 +0000 (16:45 -0700)]
Fix destroy ordering bug; add debug flag.

Bug: 19993667
Change-Id: Ib15e79a94a185843dace35fea7daaea866120f87

9 years agoRemember to tear down ASEC bind mount.
Jeff Sharkey [Mon, 13 Apr 2015 04:50:32 +0000 (21:50 -0700)]
Remember to tear down ASEC bind mount.

Bug: 19993667
Change-Id: I17093f0f9abe8762e5ee4f38ce99f9024e419dfc

9 years agoEmulated volumes above private volumes.
Jeff Sharkey [Sun, 12 Apr 2015 23:03:33 +0000 (16:03 -0700)]
Emulated volumes above private volumes.

When a private volume is mounted, create an emulated volume above it
hosted at the /media path on that device.  That emulated volume is
automatically torn down when unmounting the private volume.

Add "removed" state for volume, which signals to framework that
media has left the building, send when the volume is destroyed.

Bug: 19993667
Change-Id: I1f82b51de578ac5cfcc5d7b9a6fb44f6f25c775c

9 years agoExclusive exec() path, format after partition.
Jeff Sharkey [Thu, 9 Apr 2015 04:07:21 +0000 (21:07 -0700)]
Exclusive exec() path, format after partition.

Sadly setexeccon() is process global, so we need to carefully ensure
that all exec() are mutually exclusive to avoid transitioning into
unwanted domains.  Also, because we have several threads floating
around, we need to guard all our FDs with O_CLOEXEC.

Format all newly created volumes immediately after partitioning,
but silence all events emitted from those volumes to prevent the
framework from getting all excited.  Unify all notify events under a
single codepath to make them easy to silence.

Sent SIGINT before escalating to SIGTERM when unmounting.

Bug: 19993667
Change-Id: Idc6c806afc7919a004a93e2240b42884f6b52d6b

9 years agoam 959de99a: am f8c905e0: Merge "Add dependency on libsquashfs_utils due to fs_mgr...
Mohamad Ayyash [Tue, 7 Apr 2015 18:44:08 +0000 (18:44 +0000)]
am 959de99a: am f8c905e0: Merge "Add dependency on libsquashfs_utils due to fs_mgr dependency"

* commit '959de99af47dec9ed7a3499ce57e4097be399d7d':
  Add dependency on libsquashfs_utils due to fs_mgr dependency

9 years agoam f8c905e0: Merge "Add dependency on libsquashfs_utils due to fs_mgr dependency"
Mohamad Ayyash [Tue, 7 Apr 2015 18:13:19 +0000 (18:13 +0000)]
am f8c905e0: Merge "Add dependency on libsquashfs_utils due to fs_mgr dependency"

* commit 'f8c905e043cbeb76bf0d88dff5b5676f4e00f71c':
  Add dependency on libsquashfs_utils due to fs_mgr dependency

9 years agoMerge "Add dependency on libsquashfs_utils due to fs_mgr dependency"
Mohamad Ayyash [Tue, 7 Apr 2015 18:00:52 +0000 (18:00 +0000)]
Merge "Add dependency on libsquashfs_utils due to fs_mgr dependency"

9 years agoAdd dependency on libsquashfs_utils due to fs_mgr dependency
Mohamad Ayyash [Tue, 7 Apr 2015 01:15:53 +0000 (18:15 -0700)]
Add dependency on libsquashfs_utils due to fs_mgr dependency

Change-Id: I894d18c2086a6532194f478853a9837eb4b6fc09
Signed-off-by: Mohamad Ayyash <mkayyash@google.com>