This patch adds a new method (device_get_metadata_size) to the DDK API, which allows
the retrieval of the size of a given metadata key on a device.

Test: runtests -t driver-test
Change-Id: Ib2535b17e3e2b6b9bc225835d6ac27e288aadd52

Test: booted with qemu arm under asan
Change-Id: Ia758a685c1ad031e090e8d2e16294ff06ef765a8

Test: no functional change
Change-Id: Ia9f47f89e010fa8449326a6a3ff0c2070e03f713

TEST: No behavior change
Change-Id: I4b5a065db7e616c63710076b576872390af6cbb0

1) Remove support for scanning non-zero channels.
   virtio-scsi's config space has a field for 'max_channels',
   but the device can only encode channel 0 targets. Remove
   the max_channel loop since we can't use non-zero channels.

2) Use <= (rather than <) for max_target/max_lun comparison.
   QEMU and GCE's hypervisor disagree on the definition of the
   max_target and max_lun fields. The virtio spec is ambiguous
   as to which definition is correct. Use <= conservatively to
   not skip targets.

ZX-2314

Tested:
1) QEMU: finds target 0:1:1
./scripts/run-zircon-x64 -- -drive if=none,id=hd,file=/tmp/blk.img \
    -device virtio-scsi-pci,id=scsi  -device scsi-hd,drive=hd,scsi-id=1,lun=1

2) GCE -- finds target 1:0.

Change-Id: I684d67b8fc7449c72b14290ce88f226e00def672

The script got out of synch when the CIPD package for qemu
started using the third_party/ prefix on the package name.
Fix it up.

Test: ./scripts/download-prebuilt --no-cipd succeeds and prebuilts/downloads/qemu is current
Change-Id: I90dafd26a443318e671ba25b3fb9679c9735713f

Correct a bug in how ExecuteCommandSync called IrqRingUpdate that would
cause the scan worker thread to wait forever.

virtio ring::IrqRingUpdate will invoke a passed-in functor on descriptor
chains completed by the device. However if no descriptor chains were
completed, it would not call the functor. (I had assumed at first skim
that it would have waited instead).

Modify ExecuteCommandSync to keep invoking IrqRingUpdate with 5 mS steps
until the command completes with either an error or success. Later we will
wire up interrupt mode support, to avoid this 5 mS polling looping.

ZX-2314

Tested:
1) Booted on GCE, dm dump displayed the virtio-scsi target at 1-0.

2) Still boots on QEMU (qemu masked the error by synchronously processing
   control commands).

            [00:02.0] pid=2006 /boot/driver/bus-pci.so
               <00:02.0> pid=2884 /boot/driver/bus-pci.proxy.so
                  [virtio-scsi] pid=2884 /boot/driver/virtio.so
                     [scsi-disk-1-1] pid=2884 /boot/driver/virtio.so

Both QEMU and GCE tests need the LUN encoding fix included (see 239313) for
tests to work.

Change-Id: I9a0d8f5860f11cf1f4b900ebe3c5c4cf76a060cd

TEST=zxtest-test,zxtest-integration-test

Change-Id: I4e99d431431c3d5a09b5e640a9ec4a49a435ed1b

virtio-scsi must encode targets/LUNs into a single level LUN structure.
Originally we had a bug in the logic that did so (neglected to set
ADDRESS METHOD field). The parent commit to this one fixed the bug; add a
unit test to ensure it stays fixed.

ZX-2314

Tested: Built, run unit tests

Change-Id: I2e7468ecb7511890acba8bc17bcb854d75ca6fde

virtio-scsi requires Target/LUN pairs be encoded into an 8-byte buffer
in a particular format --

1. First byte set to 1
2. Second byte set to target
3. Third and fourth byte representing a 'single level LUN structure'

Per the SCSI Architecture Manual, the single-level format we should
be using is the flat addressing method (table 6, SAM-3 T10/02-119r0) or
(tabel 9, Flat space addressing, SAM-3 T10/02-119r0).

Bits 7-6 of the low byte should be '01'. Then we should pack the 6 MSBs
of the LUN into the low byte and the 8 LSBs into the high byte.

The prior version of this code did not set the address method (implicit
Peripheral device address method) and incorrectly mashed the LUN for
that mode.

ZX-2314

Tested: With the correct loop condition CR, booted on GCE and QEMU.
Tested that expected LUNs were found and bound to virtio.so

Change-Id: I6b4a53ea16dec0962233d143660744edd012dfc8

platform-bus.proxy.so is being bound instead of sysmem.proxy.so
(probably because it's enumerated first).

Test: boot on VIM2
Change-Id: Ib656525fe4dca8b7821d34b7773c268c760aa03a

Handles the simple case where all requests complete successfully.

Remove old batch_cb API.

Also modifies testing so we can specify when we want to set or expect
a callback. Currently those fields are equal in the tests, but will
not necessarily be when we start testing with set_error=true.

Part 2 will track queued / silently completed requests and handle
callbacks for the error cases.

ZX-932 #comment

TEST= plug in FX3 and run: fx shell usb-fwloader && runtests -t usb-test

Change-Id: I596c18b3e0edac25638214977e3668b4efccbe38

This permits userspace to query about how many debug registers
the system has.

This also removes a hack used in arm64 to pass the information to the
debugger.

ZX-3038  # Done

TEST=manual

Change-Id: I2d020b17a146b470a2f76fb97205b6516bc85cad

In order to use mt8167-gpio.cpp in unit tests, the binding for the
regular devmgr needs to be done in a different file.

Test: Boot on MT8167 ref board.
Change-Id: I3fe3aa05bfae11c707199b2b970fbe1f526f5a1e

Also removed unnecessary thread and some dead code

TEST: LED blinking and GPIO interrupt from button press on VIM2

Change-Id: I139b3dbc4edcb5aeeaf8483c71936086a3312ff3

Bug: ZX-3357 #comment
Test: no functional change
Change-Id: I042ab4263ec2bf0c155838d7ada13b17e4514224

Test: nand-loader --num-blocks 128
Change-Id: Iecca7dd201dc57674d08034b4411dd069f13eb3e

Hasn't been used in forever.

Change-Id: I2c0ffa8265c6bc5103ec332799b44607e2b7a3e1
Tested: build zircon

Using the reader policy right now through `Guard` will allow for silently violating the
thread analysis.

ZX-3085

Test: Kernel compiles, no functional changes

Change-Id: I727a87b16127413383dc2c50ea5fc83dea176344

Added ASSERT_*/EXPECT_* macros.
Added extended version to allows ASSERT*(..., fmt, args) so test errors
can capture information about the local variables that might give some
insight.
C allows EQ for comparing primitive types and ptrs.
C++ allows EQ for anything with operator== defined.

C has  no extension for printing structs. This is, custom types are
dumped as hex, primitive types (ptrs included) contents are dumped.

In C++ an overload or specialization can be provided to display custom
types of zxtest::PrintValue.

Added test for ASSERT_/EXPECT_ both in C and C++.

TEST=zxtest-test

Change-Id: I11230d0adb3d66b4c9f570604b4032441ab78fcc

It makes the syscall wrapper simpler to have SocketDispatcher use a
parameter to select between data and control plane operations rather
than separate functions.

Test: `runtests -t socket-test`
Change-Id: I9050bd50fe36173a38964b10e533c0ec9a687bc8

"RamdiskTestReleaseDuringFifoAccess" attempt to test that
the ramdisk behaves properly during termination. This behavior
is inherentely racy, as observing the test between "unbind" and
"release" or "after release" will have different results.

Rather than relying on significant implementation detail of the
ramdisk server, delete the test.

FLK-34 #done
Test: /boot/test/sys/ramdisk-test

Change-Id: I08193b929e38e092d8a0810ac1c39c82384d112f

Since the llcpp server implementation is not yet ready, for this patch,
a simple manual server is written that takes advantage of the raw llcpp
APIs and acts as the testing harness for llcpp client code.

The garnet CL used to generate the binding code is at:
https://fuchsia-review.googlesource.com/c/garnet/+/239196

TEST: /boot/test/sys/fidl-llcpp-interop
Change-Id: I960a3282b9f74506f2a5ffafaa3ece6fa6377c69

This API will be used to support hardware lights such as indicator LEDs.
The lights are identifiable by a board specific name.
A light device may implement one or more physical lights,
allowing for an array of LEDs grouped together
(for example, the LED ring on the Nexus Q music player).
Lights may be simple on/off, support a brightness value, or an RGB value.

Bug: ZX-3349

TEST: compiles when included in astro board driver.
Change-Id: Ib1b19dd27a6cf12f7171debd4a4c8715cdbe2460

Notable optimizations:
- In the callee-allocating flavor, automatically stack allocate when
message size is < 512. Otherwise use C++ new. No VLAs.
- No call to fidl_linearize when a type is completely inline.
- No call to fidl_encode when the request type is inline and has no
handles. (boring type)
- No call to fidl_decode when the response type is inline and has no
handles. (boring type)
- No reference to coding table for boring types. In the future we could
stop generating those coding tables altogether, reducing binary size.
- All empty/zero-arg requests/responses share the same struct
definition, which is essentially a fidl_message_header_t.

Also added a script to easily regenerate all the llcpp binding code.

The garnet CL used to generate the binding code is at:
https://fuchsia-review.googlesource.com/c/garnet/+/239196

TEST: /boot/test/sys/fidl-llcpp-interop
Change-Id: I4a31fe56d25504f3f76ea77c30b88a2f581a39df

fbl::atomic is being removed in lieu of std::atomic.

The underlying implementations are identical builtins, so there should
be no behavior change.

Bug: ZX-3357 #comment
Test: no functional change
Change-Id: I0538862012c3abb50e4e15e03f0a68564889091e

TEST: Added unit test to usb test driver
Change-Id: I95aaf3fad2dffc89ba536c40c355992978a5b7b2

Test: ./scripts/clang-fmt <some file> and ./scripts/clang-fmt --version
Change-Id: Ic8c8a830e3914dc0eba9b05abf4d9a8d5dac2e12

TEST: manual testing on NUC and VIM2. Tested USB audio, input and storage.
      lsusb works.

Change-Id: Ia5fea91385f3eb74191fd9a67bbec3242c45bcad

This reverts commit 68817767.

Test: revert
Change-Id: I5a54da75422f863d431db1580bd9dddafcc1fabc

All clients have been migrated to pci_map_bar_buffer.

Tested: Compiles.
Change-Id: Ide14ba02c112e47f05ec3960a074ac4f1cd2f8fc

Unittest library should also work for hosts.
Added both host lib and equivalent host tests.

The only dependency we had was zx::time, which is only used for
timer purposes. On host we fallback to std C gettimeofday.

TEST=zxtest-test,zxtest-integration

Change-Id: I937f958dc81eaab7c51b4f0bd6955517730c4139

TEST: No behavior change
Change-Id: I34fcc6d9b297ebe9633555a82d0584340b094927

Adds peek option to zx_socket_read() for both standard and control-plane
messages.

Socket benchmark tests don't show any significant change in speed.

ZX-769 # done

Test: `runtests -t socket-test`
Test: '/pkgfs/packages/zircon_benchmarks/0/test/zircon_benchmarks -p \
          --runs 1000000 --filter Socket'

Change-Id: Id312eb4889e7c2f052719980c86c75a2a1937ac4

Test: /boot/test/sys/ramdisk-test
ZX-2674 #done

Change-Id: I6ff9c767d30b4eb0a51faca1b7acc2424886f4db

Additionally, convert all clients within Zircon.

Test: /boot/test/sys/ramdisk-test
ZX-2674 #comment In Progress
ZX-2866 #done

Change-Id: Iff7751579a2a2aa496cc71da57ec002f39f15792

Additionally, redirects all usage of "ramdisk_ioctl" interfaces through
the ramdisk library (currently located within ulib/fs-management,
pending relocation). This will make it easier to convert the ioctl
interfaces to FIDL without modifying client code.

Test: /boot/test/sys/ramdisk-test
ZX-2674 #comment In Progress

Change-Id: Ia24192df10b2cb3bb5ee2a95f0b6ef647d515d69

Introducing a generic representation named TypeConstructor to capture
all areas where we reference types. This makes it possible to have a
single representation, share code around it, and most importantly, defer
type construction to a later stage in the compiler, once all constants
have been resolved.

For instance, with the example:

    const uint32 MAX_SIZE = 1024;
    struct MyStruct {
      string:MAX_SIZE desc;
      string:1024 other_desc;
    };

We must first resolve the MAX_SIZE constant to understand that the type
for both desc field, and other_desc field are the same.

Type constructors create types using type templates, which are looked up
by name. So for instance, the name "uint32" looks up a type template,
which is an instance of PrimitiveTypeTemplate, and which knows how to
instantiate a PrimitiveType representing a uint32.

All this machinery moves us closer to 1. canonicalizing types (because
their creation is now centralized, and soon easy to memoize), and 2.
introducing user-defined templates such as

    struct pair<L, R> { ....

Which would get materialized dynamically.

This also makes it natural to extend type aliasing to use any types, not
just primitives.

Test: make tools -j12 HOST_USE_ASAN=true && ./build-x64/host_tests/fidl-compiler-test
Change-Id: Id97c6751b8944b5ee04ff813d2cd0dcc47711f17

So we can use system/dev/light for light drivers

TEST: still builds
Change-Id: Ia7a6684c704390d26cedee8209f076125e7a6fec

This will make it easier to add the cb_on_error_only tests,
where we'll want to pass in which requests we want callbacks / errors on.

TEST= plugged in FX3 and ran usb-fwloader && runtests -t usb-test

Change-Id: I6078180c520ea8f61574f04316ff5801e809ece0