Function 0 is shared between all drivers using a particular card.
Currently drivers read and write to function 0 directly, but eventually
function 0 will only be accessible through one of the methods added
here.

Bug: ZX-3879
Test: runtests -t sdmmc-test
Change-Id: If1dfff0eba05a4b1d6ff255b4800ab4843b7ea78

Test: CQ

Change-Id: I2fb8dc1a01f193ec0b906ab3fe49fe5a012bb567

Change-Id: Ic753a3ebc4ab1fbc2103134514b72c6e912587ca

This CL plumbs the env_vars provided by the cmx file through the call
to fdio_spawn_vmo. For testing, the CL extends the component binary test
under //garnet/bin/appmgr/integration_tests/components to examine
enviroment variables in a manner similar to how it was previously
examining command line arguments.

Bugs: CF-723

Tests: fx run-test components_binary_tests

Change-Id: I798ce64314dd3e52538158b9ad32f5d98b50dcbb

This allows for fuchsia.io.File::GetBuffer to be called against a
memfs-backed file.

CF-748

Change-Id: I159f6cea13adcd4f1eb92fb5ab180fb5892f25cc

Many package rule invocations seem to have copy-pasta'd paths to their
binaries. This is leaning on the "stripping fixup" that finalize_manifests.py
does, adding build time and other side effects to the build. Use the
canonical path to the binaries instead, which is at the top of root_build_dir
and is already stripped.

Change-Id: I846085768206c042884dfaa0e6b0fdf780790d29

On a system where bash is not the default shell, the placement of the shbang
line meant this was invoked as a POSIX sh script, resulting in `[[` not being
found, causing a build error.

We're in general not taking dependencies on bash during the build, where this
seems to be invoked, so on the path to fixing this, I altered the script to
be POSIX sh compatible.

Change-Id: I86cf9ae112da4217c3e25402c1fe5c1b6d27dda6

Tested: modular_tests, logged in using sysUI
SCN-1018 #comment
SCN-1291 #comment

Change-Id: I0620d9d9f070e83651589370889ded488a34ddd7

This is the most basic test that program loading, dynamic linking,
and libc startup paths are working.  Nothing else would be running
if this didn't work, not even the standalone core-tests from
userboot.  But now also build this test for all the instrumented
variants, to make sure each instrumented libc build actually works.

The ubsan variants were not being tested before now and in fact are
broken.  So those variants are disabled until ZX-3986 is resolved.
When they are re-enabled, this test will sure they don't regress.

Bug: ZX-3986 #comment noop-test built for all instrumented variants
Change-Id: Iefc1c03415dbe942e7587cc7c9b01f01e6da82c3

This is a reland of f7a77a9a with the
following fix:
* The /config dir which appmgr provides to a component is not clonable,
but the /config/data dir is -- failing to clone means  /config isnt
propogated to sessionmgr, which fails a test in CI, which is good, but
not in CQ. A CQ test for this would involve faking out a /config/data,
but the testing facility for it doesn't exist today. The orginal CL
below is amended to clone /config/data instead of /config.

Original CL:

* Re-factor ModularConfigReader to parse config file during construction, and
 expose some getters instead. Have it accept a root directory to work off
 of, which makes it more testable. (i.e., fake out a directory and pass it to
 ModuleConfigReader). Provide a factory for using the incoming namespace.

* Have basemgr always pass config data to sessionmgr in its
 /config_override namespace, since appmgr enforces that /config is not
 overridable. sessionmgr uses ModularConfigReader, which prefers
 /config_override over /config when available.

* Fix modular_config_xdr to parse SessionShellMapEntry -- it wasn't
 parsing session shell before.

Test: modular_config_unittest tests that ModularConfigReader APIs
parse and return basemgr config given a config directory. Tests that session
shell URL is being parsed.

Change-Id: I737dbe897f6dc1362518bdab16dd06347ea30924

Test: CQ

Change-Id: I11c088bd872daeaa87069fbef2f9d6a96a5dda63

Needed to better understand flaky failure occuring in automation.

TESTED: no function change.
Change-Id: I8f6d8713ef0f8a9f5b0297c1ab973f4fd392ec57

This is a reland of 426805c5
NewViewTokenPair returns the old std::pair type, to allow a
soft-transition for Chromium.

Tested: view_token_pair_test; build chromium against new SDK
SCN-1389 #comment

Change-Id: Ic959f8bce804c7be99a9e7ce02e0dc59b129ca83

Change-Id: I25bfa20953af6c91b0b85ad3b09b1d53318cfabe

Notify and deactivate all discovery sessions created by the discovery
manager when it is destroyed.

BT-488 #done

Test:
Added unit test to bt-host-unittests and ran on Ava

Change-Id: If3016f16e92c4191907155b1e922fe35dc2375f7

This CL adds a 'env_vars' array of strings to the program map that
represents a list of environment variables to be passed to binary when
spawning.

Bugs: CF-723
Change-Id: Ia490819ce9155b9d04098f90e07513f48dd857b4

The hardcoded media buttons code has not
been running for some time as the generic
parser works very well. Since we are
changing the structure of media buttons
it's easier to remove this code rather
than have it rot and be non-functional.

ZX-3993 #comment

Change-Id: I2be1a9a7a3fa9c28dce2330f10e31e81d39ad5a3

Revert "[hid][buttons] Send reboot after holding volume up/down"

Revert this commit because the reboot functionality is
going to be implemented in the garnet/root_presenter
layer.

This reverts commit 1f794913.

ZX-3993 #comment

Change-Id: I42dd9ad6a2badac8e8b9a1fe88fa233dbaed332a

'fx fuzz' uses normal Fuchsia URLs to manage fuzzers.  uapp/fuzz and
ulib/fuzz-utils are now dead code.

Change-Id: I8a20570fea6b833f2a3c7713a58a6c9430ebf0c8

Change-Id: I2562e2962c89ca0cdeaed50ddd98f38682f47fa7

k cpu rdmsr <cpu> <msr> reads the value of MSR |msr| on |cpu| and prints
it to the debug console. The MSR must be readable; #GP faults are not
handled.

k cpu wrmsr <cpu> <msr> <val> writes value |val| to MSR |msr| on |cpu|.
As above, the MSR must by writable and |val| must be valid.

Many MSRs are per logical processor; some are per-physical processor;
a few are per-socket. MSR specific semantics should be considered when
accessing them via k debug commands.

Tested:
* qemu-kvm on Intel Skylake-X; wrote a sequence number of IA32_MISC_ENABLES,
  which is emulated in kvm.

  $ k cpu wrmsr 0 0x1A0 0x42
  $ k cpu wrmsr 1 0x1A0 0x43
  $ k cpu wrmsr 2 0x1A0 0x44
  $ k cpu wrmsr 3 0x1A0 0x45

* qemu without kvm; read STAR msr
  $ k cpu rdmsr 2 0xc0000081
  [00112.899] 01189:01210> CPU 2 RDMSR c0000081h val 23001000000000h
  $ k cpu rdmsr 1 0xc0000081
  [00112.899] 01189:01210> CPU 1 RDMSR c0000081h val 23001000000000h

* Eve
  (Disable Turbo Boost)
  $ k cpu rdmsr 0 0x1a0
  CPU 0 RDMSR 1a0h val 840089h
  $ k cpu wrmsr 0 0x1a0 0x4000840089
  $ k cpu rdmsr 0 0x1a0
  CPU 0 RDMSR 1a0h val 4000840089h

  (Meltdown/spectre MSR)
  $ k cpu rdmsr 0 0x10a
  [magenta screen, KERNEL PANIC!]

ZX-3359

Change-Id: I54e550e1f8ad11f94706a93b7470d952329ca934

vklatency_ is an example that we use to measure end-to-end latency from input to
vsync. This CL adds a fb based implementation that skips using scenic. It still
exposes same functionality as the skia based drawing for testing.

SCN-1376 #comment
TEST: fx shell run fuchsia-pkg://fuchsia.com/vklatency_on_fb#meta/
vklatency_on_fb.cmx

Change-Id: I15378451692e66ff0dc20cd80baefff6ee8607cd

As per discussion with infra, runtime lock violations should not
trigger the kernel panic detection. In the future a final test pass
will process the serial logs for kernel OOPS to determine any lock
violations.

Bug: IN-1145
Bug: ZX-23
Test: Trybot enable_lock_dep doesn't find kernel panics.
Change-Id: I32c6ee129add48d00bf53497d3d79ea680c0770b

This CL moves BoringSSL's command line tool out of prod and into tools.
It's a command line tool for diagnosing and manipulating TLS
connections, X.509 certificates, etc., and shouldn't be used by
production software.

Bugs: SEC-199
Change-Id: Id671c476b4818771054767e2832daab5a43aedfc

This adds the go-tuf test suite to the fuchsia test suite,
in order to make sure that go-tuf is working properly on
both on host and targets.

Note that the client test is temporarily disabled until the
python interop test is factored out into a separate package.

Bug: PKG-666 #done

Change-Id: Ic6f322b49d5ee75fb52254ccf5113456bd66339d

Setting an empty vectorptr was failing an assertion. If there's no args
we should simply let it take the default

fxr/275674 was crashing which prompted this CL. fxr/275674 has the
test that covers this change.

Change-Id: I6ff7d9cfb9995898e712fbbbf59f79346f576ede

This fixes a problem with corrupt images being generated if two
instances of this script are run simultaneously.

BUG: BLD-451 #done

TEST: built and flashed core.cleo build on Cleo
Change-Id: I411130283973af54fc07a129461d1de4897ec950

Bug: FLK-209
Test: runtests -t sdmmc-test doesn't run any tests
Change-Id: I9f93be36a75f158d4361a9b453bd0a117c614aa5

host_asan variant.
Test again when tools are updated.
Tracked by INTK-998.
This is blocking SPN-8 and SPN-9.

Change-Id: Ie56b15a6530f635a976dba9f870e3ba7a08119f9

No longer used after fxr/274961

Change-Id: Ie932b290555e70b1a0262b440b91754300110302

This cleans up a couple things with the default hardware IME:

* We recently determined text caret movement will be handled by the
application; this CL removes selection-related code.
* The IME previously had several Option<>s for various aspects of the
TextField's state, since they could only be populated after the first
update from the TextField. This refactors them out into a separate
struct of their own, so that there's only one Option<> to think about,
which we can open up at the very beginning of the input processing code.
* To support the previous point, the first OnUpdate event will call
on_first_update() instead of on_update().
* A couple functions that didn't need to be associated functions on
DefaultHardwareIme were refactored out into freestanding functions.
* get_key_mapping now returns a Keymapping enum instead of a json Value,
so that JSON deserialization is a little less spread out throughout the
code.
* Some functions that previously panicked now just return
Result<(), Error>
* All remaining unwrap()s were refactored into either `if let`
statements, or `try!`s.
* default-hardware-ime -> default_hardware_ime
* Adds tests

Change-Id: I29433df44dd3d6f2fbc211e4762d824dfe87480e

The DirectoryOpen handler will be removed along with resulting changes
to devfs in a follow up CL.

Change-Id: I37df4c6bdf2aba84575a5ab6b692b2da5f362185

Change-Id: I66ba34b62e687aab5a48b9b78f09515535cc210f

Test: Only logging change
Change-Id: Ifd2c2c5c60a12989f12b9a5146dc020cb61fe700

This CL brings 'fx fuzz' into alignement with the 'frankenbuild', in
particular it updates how to find the build directory and the zircon
host tools.

Bugs: SEC-274
Test: scripts/fuzzing/test/host_test.py
Change-Id: I3b9c36a8d79d15eb69a2a6d0133e270847b2ed49

Test: CQ

Change-Id: Ia9018276eb268edbd26eeb3e9d3545335fb3dcd0

Change-Id: I831b5b57d2ff2e3507331b239e7c283cb817812c

This CL removes the fuchsia.mdns FIDL API definition and transitions
the implementation and all clients to the new fuchsia.net.mdns API.

This change breaks the ability to generate traffic logs, the
SetVerbose method being absent from the new FIDL definition. This
capability will be restored, probably via /hub.

TEST: no behavior change, multiple regression tests in CQ

Change-Id: I6438f0a8fd316efa33a87908d41b8f9961e523d6

This change defines the disk inspector interface and also implements
basic datatypes like int32,in64, uint32,uint64, as well as uint32[] and uint64[]

Tests: Added tests to verify the behavior of the inspector library.

Change-Id: Ia87f490bb0ab52ce673785e4881f7f8caa6e53bb

This change modifies the package() template parameters for amber,
amber_tools, pkg_resolver, and pkg_cache to first invoke one or more
fuchsia_component templates and then aggregate those components into the
final packages.  This change is a no-op for now, but it will allow later
test cases to also aggregate these components into the test package.

PKG-604 #comment extract fuchsia_component templates

Change-Id: I3dae02ffd33ee410962ed622e6e664aaf309923c