The effect of this is that the VM started by compos_verify is smaller,
as well as only using 1 CPU. (It does very little, so this should be
fine.)
And that should reduce its impact on boot time.
Compilation: "--mem" "1024" "--cpus" "8"
Verification: "--mem" "1024" "--cpus" "8"
Bug: 220824234
Test: staged-apex-compile, reboot, check logs.
Change-Id: I1101965fb532abe7923ec133920d62f5eacfd108
`nix::dir::Dir` isn't Sync (required by `RwLock`), so I had to replace
it with `OwnedFd`.
Bug: 220386264
Test: atest AuthFsHostTest
Test: performance change before vs after is hard to tell
Change-Id: Ib3fd9047970a856b532234f0bebad652172e12de
Bug: 220386264
Test: Observed faster parallel file reads (at least not much slower,
possibly due to lock contention), with prototype of multithread
support.
Change-Id: I3126932f9b3c769756ec1123e4746d3d1de88a1e
If the chunk index is out of range (which can be a valid read request),
simply return 0. Previously, we had to deal with fd_server when it
returned an unexpected result. Now we don't need to handle it (the
inconsistant state may still be on Android, but we don't care here).
Bug: 221847239
Test: atest AuthFsHostTest
Change-Id: I3450f496de58f2933fb3818b42e9510c56875061
The implementation in writeZerosAtFileOffsetOnMicrodroid specifies the
offset of the output file by "skip" of "dd". But actually, "skip" is for
"if=", while "seek" is for "of=". The test expectation did not fail
coincidentally.
Bug: N/A
Test: testWriteFailedIfDetectsTampering
Change-Id: I89c9c96d250f4374e51a01c986aa1289b24ac236
A new test case is added. Besides, fix a side effect in
testWriteThroughCorrectly that truncates the backing by mistake (though
doesn't affect the purpose of the case).
Bug: 221847239
Test: atest AuthFsHostTest
Change-Id: I6b3f2b0c964c42b6bace94ff04b4f5edbc2d9efd
We should use the system properties instead (check for capability not
implementation). And we should use the new utilities to do so.
Bug: 218276733
Test: m ComposHostTestCases AuthFsHostTest MicrodroidHostTestCases
Test: atest ComposHostTestCases
Change-Id: I5db819dd874487310a40cb6736f873d3b6273ecd
This removes schema-mapped Java classes which are duplication with
Rust's mapping type.
Bug: 221368644
Test: atest MicdrodroidHostTestCases
Change-Id: I8d2fe9ac21a887543adb05298ddaaae0b413d490
pvmfw.img
Built here: ab/8231605
The change is generated with prebuilt drop tool.
Bug: 217161634
Test: boot a VM
Change-Id: I31c7ce92c01fc7afae86a9b3b39bda7aead147fd
We've found signing can be run where truncate is not available.
Error:
No such file or directory: 'truncate'
Use python's truncate() instead.
Bug: 193504286
Test: atest sign_virt_apex_test
Change-Id: I9c8a78ba381274b406ba509f5f3f7ddd049bd60f
We run VS at early boot (via compos_verify), and we want to see its
logs in case something interesting happens. That seems to require
using the sytem log.
Bug: 215419883
Test: Run staged-apex-compile, reboot, see logs
Change-Id: I24c55196b5403df3057c690691f0f585c4a8f2fc
VM config can now specify the requirement on the virtual platform
version. At runtime, the requirement is matched against the actual
virtual platform version that crosvm implements. If they don't match,
the VM can't be created. The version format follows SemVer, allowing us
to express backwards compatible and incompatible changes in the future.
Bug: 193504487
Test: atest VirtualizationTestCases
Change-Id: I23d370081e10399502178b9cfe8a46b05addf186
The test needs root to run and does not add much over
MicrodroidHostTestCases. Remove it from the CTS suite.
Bug: 218750257
Test: run CTS
Change-Id: I084e43058da5ec05c330c6f08e86a1f0b0a1b1e5
On Pixel devices, adbd/statsd apexes have different names, which causes
`pm path` to fail.
Instead, embed test-purpose APEXes as data and use them for testing.
This can also provide isolation for the testing environment.
Bug: 221024307
Test: atest MicrodroidHostTestCases (on internal branches)
Change-Id: I4a71ea662e789d48bdbdf2008bbdd666778a72fe
Bootloader fails when its embedded key doesn't match with vbmeta's
signing key.
Note that VM itself doesn't shut down even when boot fails for now.
It should shut down the VM just like pvmfw does.
Bug: 218934597
Test: atest MicrodroidHostTestCases
Change-Id: I8bfef33d3aafe72f672c035fe796d8636bc95e1c
Treehugger Robot
Alan Stokes
Victor Hsieh
Inseob Kim
Jooyung Han
Charisee Chiw
Andrew Walbran
Pierre-Clément Tosi
Jiyong Park
David Brazdil