The code was not written with the expectation that global dtors run, and
executing them currently results in ASAN violations.

Change-Id: Ib9e709483b30a8308af59009c0cbb22fec1fd2e1

This does not provide the actual implementation in devmgr, just the
plumbing to send the message from the driver to devmgr.

Issue: ZX-2648

Change-Id: I9c3160d86a35c19313b4cf7450a8e3e7ca4140f0

This adds a simple set of sl4f endpoints for controlling audio
on a DUT.  It currently supports Injecting audio into the input
path and Saving the output from output path.

Test: PutInputAudio endpoint
 {
   echo -n '{"id": "", "method": "audio_facade.PutInputAudio","params": "'
   cat test.wav | base64 -w 0
   echo -n '"}'
 } > payload.json
 curl -X GET [`buildtools/linux-x64/dev_finder list`] -d "@payload.json";

Test: GetOutputAudio endpoint
 curl -X GET [`buildtools/linux-x64/dev_finder list`] \
      -d '{"id": "", "method": "audio_facade.GetOutputAudio",
          "params": null}' \
 | python3 -c "import sys, json; print(json.load(sys.stdin)['result'])" \
 | base64 -d > audio_capture.wav

Change-Id: I0c52f1bd7723b346771308731722bcc57a668b04

Allows trace_cout to be used as a tracing mechanism for host-side
Overnet.

Change-Id: Ia9dfaf3325c79f759f33973d9f56f63b32c0e0e0

Change-Id: I82ba83e4cb132ae5f5b21fe83f59280afdb34825

This API is not yet implemented.

Issue: ZX-2648

Change-Id: I81a17516e5e2dc8cd746f94170dc214ce1ecffb8

Patch 01d639cf changed boot arg handling
by saving them to a map rather than reading them out of the environment,
with the effect that now later boot args override earlier ones.

Given this new behavior, this CL moves the default TERM handling to the
beginning of the boot args so that the caller can override it.

Test: `scripts/run-zircon-x64 -Cb -c "TERM=dumb"` then `echo $TERM`
      now properly shows "dumb"

Change-Id: Id4f78bb159a65a16eaa64172154ca15372cb20b7

BUG: FLK-89 #comment

- Relax global setup timeouts and helper_netstack_cfg timeout.
- Logs in reported flake suggest that the timeout is too strict.
- Adds more verbosity to helper_netstack_cfg so it's easier to diagnose
where it got stuck.

Change-Id: I9975440fe3c9fb0d30492f5671785a76f367f37f

Explicitly test that new boot arguments override old boot arguments.

ZX-3385

Change-Id: I0a648773d1f55f9052eb7dca060dea4fa84257ee

Remove unused constants. Convert #defines into constexpr. Convert Macros
to templates.

Test: iochk, CQ.
Change-Id: I7526d28bc166ed6eaf7eed8b77bc211569c59709

This directive includes a lot of undesired things like
allow(unused_imports) and, more improtantly, allow(unused_results).

For temporarily allowing dead code, #[allow(dead_code)] can be used on
per-module basis, although it's likely unnecessary for a library crate.

Change-Id: I74a6de50c83de08ec3af9bad91bff617e585498e

vendor/ansi_term already existed. This CL brings in the usability.
Cargo.toml is manually modified.
Cargo.lock is done by fx update-rustc-third-party

Test: fx full-build
Change-Id: I7b89cd1a33f49cfa7f3b1bdec989cc5eae445255

Test: Ran carnelian examples, terminal, and voila
SCN-900 #comment
SCN-905 #comment

Change-Id: I92cbbb72b7b258095727c9f5d76114ee067e92ff

Also add DID definitions for each of the three SD/eMMC blocks on the
chip. This is necessary for the SD and eMMC devices to be usable at the
same time.

Test: iochk SD card and eMMC partitions.
Change-Id: I67565bdd234ebd0210cd7adb59c09bac91c6d8e4

It doesn't need it.

MF-261 #comment
TEST: ran modular integration tests which exercise dev base shell

Change-Id: Ib9ab99e22e034b25ba56b847b25c11e105712bb1

Change-Id: I3618b2cdc1bb74a086b259dadb00d93d0b3e06a2

- We previously had an assertion that could be triggered by a remote
  host if they somehow got us to process an IP packet with a 4GB IP
  header
- Since we don't care to support such an extreme edge case, and such
  a case can almost certainly never happen in practice anyway, we
  simply allow our logic to be wrong in that case, and add a comment
  explaining this

Change-Id: I0fce32fee9b48566a9937ea9060efdde5375e26e

We switched "dm poweroff" to use a new service provided by the devmgr
instead of /dev/dmctl but forgot to bring the pwrbtn-monitor along for
the ride.

Issue: ZX-3403 # handle reboot,shutdown,suspend.
Test: Ran on eve, powerbutton shutdown the system.
Change-Id: I3c965dedb9e82f6bb1567ee1823a0d8b005e321d

Change-Id: Ib888c1c2e73ebb415f807bc3ff49b049646993ff

In particular, remove the assumption that futexes will be released
from their wait_queue in the same order that they joined the
wait_queue.  This used to be true when the kernel maintained it's own
independent list of thread who were waiting in a futex along with a
dedicated wait_queue for each waiting futex, but AFAICT we were not
making this guarantee in the documentation.

Furthermore, we probably don't want to make this guarantee.
Particularly as we add the concept of priority inheritance to Zircon
futexes, we probably would like to make no promises here so as to
allow the scheduler the ability to control which futexes are selected
for wake/requeue operations when the time comes.

In addition to this, generally update the tests to conform a bit more
to the modern Zircon C++ style.  Specifically, make more use of
fbl::atomics, fbl::AutoCalls, libzx, and other helpers that we have
developed since these tests were originally written.

Test: Build x64 QEMU and run updated tests.
Change-Id: I9f837bdceba2badb1be7fe5e9ca9da18f052a6d1

Change-Id: Idfb5020b35ee87ec2618c88821b98b08e5cba050

Bug: ZX-2845
Test: CQ
Change-Id: Iae87736f1b1f188db0bf62331aad9f9f348c81af

Bug: ZX-2845
Test: CQ
Change-Id: I078cb7306af59702c3f94484113723873ef9f749

- Add the ability to send ICMPv6 "packet too big" messages in
  response to attempting to forward a packet over a link with
  an MTU too small for the packet
- Store MTUs as u32s rather than usizes (ICMPv6 "packet too big"
  messages encode the MTU in a 32-bit field)

Test: None; we don't have IPv6 supported for test dummies yet;
      once we do, we will add tests for this and other ICMPv6
      functionality

Change-Id: I42b24f8fedbce3c97ab6376e86da949a08ed2dff

Missed in 7ab9e564 and d76aea86.

Change-Id: I82846fa646d9df2f03f5564c5f1acab5d8d0c39b

Regular comments should be flow-through (///), re-organising the content
to make that clearer, and speak about non flow-through comments as the
exception.

Specifying that comment widths should be limited to 80 characters.

Change-Id: I80ba31dbc62272f3a652bac3902c83d64e05414d

Apparently the ath10k requires QoS is enabled to send out packets
at 11n+ rates. In this CL, the MLME will mark the client supports
QoS and let the driver/firmware to decide what to do.

BUGS=WLAN-767,WLAN-108
TESTED=With my Toulouse (as AP) and Ava/Pixel 2 (as client).
       The AP can send out packets at 11n rates.
       Also regressed the ralink and broadcom drivers to ensure
       they are not broken.

Change-Id: Id75ce89c7717e759d8e4d405e94ae27d1ac38335

Change-Id: I9fb9d073f1a0c852c9e6019e276698c5731c67c4

Allows host side Overnet to provide a different encoding/decoding path
to support proxied handles.

Change-Id: I65735c3db612dc1bfa27d1a83ef6a0108b0db4f2

- Replace uses of ! (the never type) with never::Never

Change-Id: I0eca8a86cfabc0acaa8fdf50759fc9992b6f7b54

Adding a method to make it possible for accessibility manager to send a
request to a client to take  certain actions on a node.

Testing:
  * Build: Passed

MI4-1736 #comment

Change-Id: I4ccc80fa8915b91ef1eb5f51c03a5d28f6ec8ac3

- never_type feature was used as a placeholder in a previous commit,
  and accidentally committed
- Replace with the never::Never type

Change-Id: I5c4562662e04e14ed172ec62c258dc3bdd581fb4

There were a number of things rustfmt complained about in here.

TEST=fx rustfmt .:bin from garnet/bin/sl4f no longer changes anything

Change-Id: Iaf009b2b7ace22003ca1b705d1b4575984c178af

Change-Id: Iba5dd182b9b2e063b69fead6ee577e41e226a540

New features:
 - Generate Tricium proto compatible JSON output.
 - Consume a list of paths to JSON IR that's produced by the build system.

Given a couple of built Fuchsia trees:
  ./scripts/difl --before-files=out/fidl-json-bak/all_fidl_json.txt \
                 --after-files=out/x64/all_fidl_json.txt \
                 --format=tricium
will describe the ABI changes between the two in a Tricium-friendly JSON
format.

Given an unbuilt tree that has run `gn gen`:
  ninja -C out/x64 $(cat out/x64/all_fidl_json.txt)
will just build the FIDL IR. Note: It'll expect to already have `fidlc`
from a Zircon build.

Change-Id: I4d56706ba4740bd24df76f7bee5de41d0e8fa306

Change-Id: I52fa19a231fc37adcb01308868ae57038fd58def

TESTED=Ran "ifconfig bridge eth" and checked output.

Pair: stijlist@google.com

NET-2118 #done

Change-Id: Ie0dbf9fafa2fccc919711260458ca081c0563a0d

- Separate element splitting from parsing. This allows e.g. finding a
  single specific element in the frame without parsing others.
  This also avoid unnecessary "serialization of state" (switching on
  element id twice)
- Make a universal write_ie!() helper macro that makes element
  serialization much easier and less repetitive. This is similar to the
  current C++ template-based design.
- Split the module into several files to keep it more organized.
- Introduce a strong type alias for element ID.
- Make element parsing as zerocopy as possible. E.g., TIM header is not
  copied when parsed. This also allows in-place modification of fields.
- Make sure all parsing works over a generic ByteSlice and is not
  limited to immutable slices.
- Explicit error types for frame parsing and writing instead of the
  generic failure::Error. This allows the client code to recover from
  specific errors safely and explicitly.
- Helper macros for frame validation during parsing and writing.

Change-Id: I88262817cc51996c9575e0bf99ab729e3bf18f68

This CL also includes new rust formatting changes.

BT-763
Test: act.py -c <config> -tc FuchsiaCmdLineTest

Change-Id: I12a78b507693fea7a259c307baeb9516acd6282f

This CL allows keyboard input for clear event happening with "c" key.

Change-Id: I081bec5f28814625fddc1bcc7b172e0ee7547428