When realm's job dies, that realm becomes useless because it cannot
create any more components aor realms inside of it, so better to kill it
and lets realm owner know that realm was killed so that they can exit
gracefully or create a new one.

This will also complete bug CF-706 because now we have checks in place
which ensure that appmgr doesn't crashdue to closed handles  when there
is a oom.

CF-706 #done

TEST=fx run-test appmgr_integration_tests

Change-Id: I4a06bc73d76fe44a075d63fc3f0f762a94a19a14

This code was all written from scratch for Fuchsia.
Move it to system/ulib/c to make its provenance more clear.

Test: no effect on the build
Change-Id: I6b4105157c9233ad3b7ec41124c35cbf144c2ae0

This unifies the API for C++, Dart, and Rust better.  All 3 language
bindings are pretty idiomatic now.

Tested: view_token_pair_test
SCN-1214 #comment

Change-Id: I00bcbde5ff64c747ce427764cdfffdf6530e1295

This is the start of fdio conversion from C to LLCPP.

TEST: Compiles
Change-Id: Ia546d5199d9b73677cb30ae38b445f45225fea34

Implements fx build tools/fidlgen_llcpp_zircon:validate.
In addition, adds the validation pass to the fuchsia build.

Test: Use fuchsia-mem:llcpp somewhere, then change it. Observe that the
build fails with an error prompting the user to re-run bindings
generation.

Change-Id: I376516c794120fa5faaaf4e2afaffc0283604d09

Change-Id: I24a609e37a42c956514076b330b924464e73ed41

ES-198 #done

Change-Id: I5dcdec7f10067c90d0e8fc18d17069f26bbfeb59

TEST: tried importing the generated fuchsia.mem in some test and it
builds.

Bug: ZX-3874

Change-Id: Ibb963f4581e0f9fc1bfca862d36b6fb0a76335c0

This support is no longer being used.
All drivers are now binding to the generic I2C either directly
or as a composite device.

TEST: Touch screen works on astro and cleo. runtests -t platform-bus-test.
Change-Id: If189943cdabfebc2fab2df1010881daf458ada6b

When the number of interfaces grows, the vector containing the interfaces
grows with it.  This involves the interface being moved.

When you generate an interface for the first time, each interface method is
given a pointer to the enclosing interface.

When the interface vector is resized, the interface method's pointer is
dangling.

This CL fixes the issue in two ways:

- The move constructor for the interface now creates an updated interface
method instead of reusing the old one.

- Interface vector initialization now reserves enough space for all of the
interfaces, so that moves will no longer be necessary.

Change-Id: Ifa4c51f21bb3dc194695d5af5029a46b862ae907

* there is no need to run the underlying delayed task if the log collection already completed

TESTED=manually removed the complete_ok() in Done() and checked the timeout callback is no longer called

Change-Id: Ic844d6ca95ebe65d31651b080e4e26a4cece7870

This implements half the FIDL generation for LLCPP bindings.  It
can't be done fully the straightforward way until we can build the
fidlgen tool (Go).  Instead, this integrates it halfway by doing the
JSON generation part.  Then instead of running fidlgen, it produces
a build API module (JSON) describing how it would have run fidlgen,
and expects to find the fidlgen output files in the source tree.
Later integration will take that information and do the regeneration
and/or check that it doesn't need to be done, so that the
higher-level build can verify what the Zircon build alone can't yet.

Bug: BLD-442 #comment fidl_library() support for llcpp
Bug: BLD-353 #comment Switchback for FIDL LLCPP bindings before unification
Change-Id: Ied2e9972da1c1bb4957e29f98c721d96d8839d97

1. Because Google app doesn't require callback interface between the FTL and
   the NDM, removed it to increase clarity and decrease lines of code.
2. Changed some local variables from 'ftl' to 'ftl_cfg' to better differentiate
   between control block handles and temporary configuration structures.
3. Removed ENOMEM definition from 'kernel.h'.
4. Removed FsError() calls from routines in 'fsmem.c' since their callers
   invoke FsError2() if there is an error.
5. Introduced 'ftl_pri.h' (for internal use) and 'ftl_pub.h' (for driver).
   Following check-ins will consolidate other headers into those files.

Change-Id: I6c27dfd24755c5fda1e0cb5ba09e8707ce31b653

Create an abstract "BlockDeviceInterface" class, and
add unit tests for it.

Change-Id: Ie61268a11a217a27e23fabfbdd5873d01e734a8a

FLK-187
FLK-188

Change-Id: I25a5a2c72bd5a5348953e315142bc2bdc8fff610

This changelist encapsulates the SetUIService into a struct/impl.
This structure will facilitate future changes where state and logical
units will need to exist across channels.

Test: fx run-test setui_service_tests -- --test
Bug: SU-167
Change-Id: I850d75321a7d9f830bc611ba4ffc3bfcb6680534

Renamed session update function and moved CommandContext creation to
be assigned the trace ID, rather than guess the next ID.

Tested:
- updated unittests
- ran astro

Change-Id: I421ffe839c9add7440361209cc881c66c80e0843

Previously, it was possible for a hot-looping task to be
created as a result of a stale PEER_CLOSED signal arriving
for a new, still-open channel. This change prevents this
by ensuring that the `PEER_CLOSED` signal gets reset and
re-waited in the event that a read or write occurs without
resulting in PEER_CLOSED.

This change fixes the issue for zx::Fifo and zx::Channel,
and a followup change will be opened for zx::Socket.

Change-Id: I0220b084320158d20d624b75c4267b0f87e6687f

Instead of loading bundled defaults from the system_image package
start requesting the config-data feature and also looking for the
settings from the config-data path.

MTWN-262 #done

TEST=none

Change-Id: Ia1cd55f6c8bb1f8ae5546d651b50492642f867c9

Previously this library only supported connecting to Cobalt via the
older, heavier technique of passing a VMO containing the registry bytes.
Now it also supports connecting via the newer, lighter technique of
passing just the project name.

Test:
I added a new test case to cobalt_logger_unittest.cc

Change-Id: Id90ff6fc67faf73d7b19d9dc8a9691adee48de49

Update the ARP table on receiving gratuitous ARP requests and responses,
as specified in RFC 2002.

NET-2102 #done

Change-Id: I8f78200deb2f36f5cfff46bfbd9baec13f2d161f

Cpp Style (anonymous namespace for tests, and CamelCase for function
names).
Use zx type safe classes instead of raw zx_handle_t.
Prevent leak of handles by early release of a[0] on failure.

TEST=core/handle-transfer-test

Change-Id: Ic9ec5346d70afb9846f233c814ad709a3b35addd

BUG: FLK-200 #done

- Test case "TestSucceedsBeforeTimeoutFires" was causing flake due to
low timeout value (5s). Bumped value to 60s to prevent false negatives
in timeout test case.

Change-Id: Id7a999a3025291db12c7c39a3b1c20ced154e89c

Add basic tests that exercise non-blocking and blocking calls from dpc_queue.
Check that a DPC runs on the same CPU that it was queued on in the absence of
hotplug. Also check queueing many DPCs and seeing that they all execute.

Tested:
* k ut dpc_tests on qemu x86, qemu-kvm x86, qemu ARM
Ex (qemu ARM):
$ k ut dpc_tests
[00010.025] 00000:00000> dpc_tests : Running 2 tests...
[00010.030] 00000:00000>   basic test of dpc_queue    : PASSED (3393392 nSec)
[00010.037] 00000:00000>   test queueing lots of DPCs : PASSED (6933168 nSec)
[00010.037] 00000:00000> dpc_tests : All tests passed (2/2) in 11158208 nSec
[00010.041] 01188:01209>
[00010.041] 01188:01209> SUMMARY: Ran 1 test case: 0 failed
$

Change-Id: I405c9eaf99e9f6050378293889d417e97a54ab81

Previously we were removing jobs from the tree then marking them as
dead, but since removing from the tree can trigger the parent to die,
this meant that parent jobs were dying before children.

This CL switches the order so that jobs now correctly die from the
bottom of the tree up. This means that there is a brief window after a
job dies where it still exists and could be retrieved from the job tree,
but it's OK to have a handle or operate on a dead job, most functions
will just return ZX_ERR_BAD_STATE.

The reason this only manifested as test flake is because the job killing
thread would usually be allowed to complete before zx_object_wait_many()
returned in the test thread, but every once in a while
zx_object_wait_many() would fire quickly enough to catch jobs in the bad
state.

FLK-189

Test: new DEBUG_ASSERT() fired in old code but doesn't now
Test: `while runtests -t job-test; do :; done` no longer fails
      after a few minutes (QEMU ARM)
Change-Id: I38119d401c2f6e3310882a58c3d9b52e684d3d34

The original implementation offered only two options for zxcrypt's
key-source policy: always use the TEE, or always use a null key.

For a variety of reasons, including avoiding immediately losing data in
an OTA moving from the null keysource to the TEE keysource, we'd like to
be able to make a smoother transition, where we might try to use the TEE
if possible but be willing to fall back to the null key.  This patchset
expands our options by also allowing a policy for "opportunistically try
the TEE, but fall back to the null key if that fails for any reason" and
"require the TEE for new volumes, but still try the null key to unseal
existing volumes".  The former is safe to deploy even if we're waiting
on bootloader changes to land to enable the keysafe TA, and the latter
helps start the clock on devices which aren't using TEE-backed keys.

After spending sufficiently long with a board configured as
"tee-transitional", we can safely cut the config value over to "tee"
(required) since most devices will have been paved with a new zxcrypt
volume, which will be using a TEE-backed key.

Tests: With a modified bootloader and the keysafe TA included in the
system image, I OTA'd an Astro from "null" to "tee-opportunistic" to
"tee-transitional" without data loss.  I then repaved the Astro under
"tee-opportunistic", then OTA'd to "tee-transitional" and then to "tee",
and observed that each came up and successfully unlocked the /data
volume.  I also added unit test coverage.

SEC-270 #comment

Change-Id: I0fb95e5322468da4e13004e21569f44385d8d8c8

Test=Included unit tests

Change-Id: Ibeb2bec199ff4d43e4de4b8c69e86ac6b7930973

This makes the CobaltEvent interface more uniform with the existing
Log* methods. The only unsupported Log* method is LogCustomEvent.

Test: Add some tests for logging CobaltEvent (including new variant).
Change-Id: I28720826d7a552141d5e0ea64cdfe4e533261262

Test: fx set core.x64 --with bundles:buildbot passes.

Change-Id: I8783140a4380b9777bf41eb284a016c9fdf8a73b

custom .clang-format is removed for drivers so conform to the fuchsia
wide .clang-format

Test: Existing tests provides sufficent coverage. No behavior change.
Change-Id: Ic101de799d831afa545a171fe0ed0a15d7efef21

DPCs are a mechanism to queue work in the zircon kernel to run later from
thread context with interrupts enabled; comment the contract of DPCs more
completely.

* DPCs are run with interrupts enabled
* DPCs are queued to the current CPU and run there except in the presence of
  offlining/hotunplugging a CPU
* DPCs will never move CPUs while executing

Change-Id: Ib8d7d73cf6f8df99f49f632434387a3fd71f7f30

The receive_arp_response test function is not generic enough for the
upcoming gratuitous ARP implementation. This is necessary to send both
request and response ARP packets without reimplementing the same logic
and assertions.

The solution to this was to rename receive_arp_response to
send_arp_packet, which allows for specifying the operation of the ARP
packet (request or response). The function arguments were moved around
to mirror receive_arp_packet.

Change-Id: I815ff437656cdddd0f51f67983df24a95f1e124e

We request that the virtual audio driver notify us every 50ms. If we
don't receive notifications for more than 150ms, it's a sign that the
system is under too much load and tests using audio are likely to have
issues.

Test: Built with threshold set to 40ms to confirm that it fires, then
      swapped it back to 150ms.
Change-Id: I800ae1c3937c2f6ba03e1baea347c32393b11aca

1. Streamlined memory allocations.
2. Fixed ftlmcFlushPage() and ftlmcInCache() to update LRU ordering.

Change-Id: Iedccf0e88fc5abe56db2994a2d46223d010a6b80

Test: time-test

Bug: ZX-3741 #done

Change-Id: I37b0a2e3947ba29b63fa4b1b1f066a9e88231497

The bug in the underlying infrastructure has been fixed.

LE-624 #done

Change-Id: Ib65ef2cc3e98275a3481c93b2a7fed52119b83ee

* //sdk:modular_testing is an SDK that provides testing utilities for
  writing app & integration tests that run under modular. This CL
  bootstraps this SDK with just the runtime necessary to use the
  fuchsia.modular.testing.TestHarness.

* sdk_image.gni now contains the modular_test_harness fuchsia package,
  which is required by //sdk:modular_testing

MF-328

Test: The c++ gtest fixture is tested by test_harness_fixture_test
Change-Id: I3e4944ac25ecf4cf14960d6386ce64e8a2a935d6

Test: No functional changes.
Change-Id: I37e6803d14baa7cd94816bf05686ab5c89cd8be4

This CL contains most of the functionality need to reach feature parity
with the existing implementation of oldhttp. The one notable missing
feature is that this implementation doesn't actually return the body of
the HTTP response.

Change-Id: I94c15a9fac742dde9d058f7c22ab4628ec6d2366

Parts of the build that export data for scenic will start putting
it in the config-data package. In preparation for this, update
scenic to look in both the /system/data path and the config-data
path.

TEST=CQ

SCN-1356 #comment

Change-Id: I01561154ea51fe69e61e55f2b50239cf8e6cd506