In case of GSI on DSU, /data is mounted on the following stack:
userdata -> userdata_gsi -> actual block device, and fstab returned from
ReadDefaultFstab() function will say that /data should be
mounted onto userdata_gsi. This configuration wasn't supported.
To support it, fs_mgr_get_mounted_entry_for_userdata now unwinds the
dm-device stack /data is mounted onto, and for each entry in fstab
resolves corresponding block device (in case of logical partitions by
asking DeviceMapper to resolve name -> path, and for other cases by
reading a symlink), and then checking whenever resolved block device
belongs to the unwinded stack.
Test: on both normal boot & DSU
Test: atest CtsUserspaceRebootHostSideTestCases
Test: atest CtsFsMgrTestCases
Bug: 148612148
Change-Id: I1bb3e64ae92d56de905b4f7f1997b7499b101b0f
Merged-In: I1bb3e64ae92d56de905b4f7f1997b7499b101b0f
(cherry picked from commit f495fe798b)
... so that an operation can be skipped partially. For example, if
an operation contains blocks:
563412 -> 123456
... then optimized operation is:
5612 -> 1256
Test: update_engine_unittests
Test: apply incremental OTA
Bug: 148623880
In an experiment, this reduces CoW size of an incremental update
package by 200MB (out of 700MB).
Change-Id: I86ca23fd589ddbc84c81318283b5f4e71782a759
GetGsiService() is defined in libgsid, let libfs_mgr_binder link to
libgsid instead of duplicating the logic.
Test: Build and boot
Change-Id: I28d0c92b432a53770a3bdc5f4c5cd91e3466151c
This reverts commit faa49d1d00.
Reason for revert: Callstack should not be logged.
Test: none
Bug: 148818798
Change-Id: Ie5506ae0c7408e255a464b2f403d0a47d272229e
Now that update_engine is responsible for initiating the
merge, InitiateMergeAndWait function becomes useless.
Bug: 147696014
Test: libsnapshot_test
Change-Id: I5473dc543ca8ac2fd31f597720847b02d0e5e33d
The rollback indicator is created during first stage init and
is unlabeled. Need to restorecon it properly so that it can
be properly deleted by update_engine later.
Bug: 149956852
Bug: 150907679
Test: apply OTA, reboot, look at serial console
Change-Id: Ie954023b557865691fec2ae378145596bf1973a9
Logic of unwinding dm-device stack to figure out what entry was used to
mount userdata turned out to be a little bit more involved, and it
shouldn't be part of libfstab
This CL just moves code around and cleans API a little bit, actual fix
will be in the follow-up CL.
Test: atest CtsFsMgrTest
Test: atest CtsUserspaceRebootHostSideTestCases
Bug: 148612148
Change-Id: If0f8a765dba63adb0e6a711eb81fffdcabea3480
Merged-In: If0f8a765dba63adb0e6a711eb81fffdcabea3480
(cherry picked from commit 8f50cfc28d)
The file accessing API in liblp don't work on win32. Add the O_BINARY
to correct the file stream access.
This patch follow the AOSP change 829979.
Change-Id: I97180e2f07cacf109a86593b6a12975597552053
Signed-off-by: Haoran.Wang <elven.wang@nxp.com>
Change API for SnapshotMergeStats so that it is easier
for update_engine to use.
Test: apply OTA and look at stats
Bug: 147696014
Bug: 138817833
Change-Id: Ie4036ac7382102c00f0761f443d78e00b9e585d5
Now that update_engine is responsible for initiating the
merge, WaitForMerge function becomes useless.
Bug: 147696014
Test: compiles
Change-Id: I3779bdf9dd8d1716238e21d09acf0499e3fc90a3
In CheckMergeState / HandleCancelledUpdate, if removing update state
fails, just return the previous state. It used to return Cancelled,
and the error goes unnoticed.
Test: pass
Bug: 147696014
Change-Id: I9cb3d20c5c886afa1913740c903eaad08f0cc041
If updated then immediately flashed target slot at unverified
stage, update_engine attempts to call RemoveAllSnapshots, but
it used to fail because unmapping fails. In reality, these devices
are dm-linear targets, so unmapping them is optional.
- Introduce a GetSnapshotFlashingStatus function that expose more
information than AreAllSnapshotsCancelled. We can use the returned
map<string, bool> to determine whether a partition is flashed or
not.
- Introduce ShouldUnmapDeleteSnapshot which determines whether
unmapping / deleteting a snapshot is needed in RemoveAllSnapshots.
Test: apply OTA, flash target slot, then inspect logs
Bug: 147696014
Change-Id: I0853d1e213566af2a3401e46fac7d9586cee7aaf
If rollback is detected in ProcessUpdateState, call
RemoveUpdateState and return UpdateState::Cancelled. Now that
update_engine is reponsible for initiating the merge, it can react to
this state and clean up markers appropriately.
Test: libsnapshot_test
Test: apply OTA, manually rollback (by setting the active slot), then
inspect /metadata/ota as well as /data/misc/update_engine/prefs.
Bug: 147696014
Change-Id: Ibfee11fb50e4f4fb7c6cf02b4921b35e77b8f5a5
Init skips first stage mount entirely in recovery, so
the check is useless. Also, __ANDROID_RECOVERY__ is always
set for first stage init.
Also add logs if rollback indicator is set.
Fixes: 149956852
Test: apply OTA, manually rollback, reboot, see rollback indicator
Change-Id: Iac967baf4b51ea5ce9f6eb962a33cb7ee5819c1d
Change callback type of ProcessUpdateState to bool() so that, when callback
returns false, immediately returns Merging.
Test: libsnapshot_test
Bug: 147696014
Change-Id: I9dcb8e1658b95216c0f1991e2e4c1ea2e7e7b2e5
These functions are used directly in update_engine now that
it is responsible for initiating the merge. update_engine has
its own message loop, so WaitForMerge / InitiateMergeAndWait
becomes useless and are deprecated.
Bug: 147696014
Test: compiles
Change-Id: Id8cb14f5f18b4893e2fc3d0d2b374a867a220e6f
Add an optional prolog arg (function<bool()>) that is invoked
before snapshots are deleted and update state set to none.
This allows update_engine to delete markers before deleting snapshots
to avoid depending on the erroneous markers. Otherwise, if update_engine
delete markers after libsnapshot deletes update states, the device could
technically get into a state where update_engine thinks the update has
been applied, but snapshots are gone.
Bug: 147696014
Test: libsnapshot_test
Change-Id: I71bfc04a81ea4f94b3072558be50d2f80565113e
update_engine is now responsible for initiating the merge. snapshotctl
becomes a debugging tool for libsnapshot now.
Bug: 147696014
Test: libsnapshot_test
Change-Id: Ia2527a35e0c0f0789dbe5c477e174663ef406903
Enable the --report flag in init rc script to collect and send snapshot
merge statistics after OTA.
Bug: 138817833
Test: statsd_testdrive
Change-Id: Ie32a2c6d7d1671ca2b1846c6a8d33cea2ab22a4c
Signed-off-by: Alessio Balsini <balsini@google.com>
Replace the original on demand start mechanism with the new dynamic
AIDL service infrastructure to resolve a possible race condition.
Bug: 149130673
Test: gsi_tool status
Change-Id: Ia5f32579a8dcf62d700d974c7f4e3c65647f3b8b
Add the --report flag to spshotctl to submit Android metrics on the final state
of the snapshot merge operation, the duration if the merge is completed in a
single session, or the number of reboots that happened before completing the
merge.
Bug: 138817833
Bug: 148138124
Test: m + OTA + statsd_testdrive
Change-Id: Id657f33963bb9926553fa6eeb1721a5229f11e87
Signed-off-by: Alessio Balsini <balsini@google.com>
Merged-In: Id657f33963bb9926553fa6eeb1721a5229f11e87
WaitForFile/WaitForDeletedFile both early return true if an error like
EPERM occurs. This was intentional because the code was modeled off
earlier fs_mgr code, but it makes libdm inherently racy if sepolicy is
not configured correctly. It's better to have these result in explicit
and consistent failures.
Bug: 148103327
Test: fastboot flashall
Change-Id: I0c78818962e1db91b556e523c418db28f7d78fae
Merged-In: I0c78818962e1db91b556e523c418db28f7d78fae
Some tests were temporarily disabled due to errors in Cuttlefish.
Now that the issues were fixed, the tests can be put back in presubmit.
Bug: 148889015
Test: vts_libsnapshot_test (Cuttlefish + Pixel 4)
Signed-off-by: Alessio Balsini <balsini@google.com>
Change-Id: Idcb2b7831b2183b2d734bdf8821fd34746f2beee
EXT4_FEATURE_INCOMPAT_CASEFOLD is a flag in s_feature_incompat, not in
s_feature_ro_compat.
Bug: 138322712
Test: Treehugger
Change-Id: I07bb1d2b818d423196d8ceebda8900a8adbb19e9
fs-verity is required for new devices launched with R. This change
remove a manual setup for vendors going forward. The original fs mgr
flag still allows old devices to opt in, which can only to be done
manually because of kernel dependency.
Test: build
Bug: 150034150
Change-Id: I152b63d7889153d41f29677f72074afb1881b65d
