Method ordinals will soon be not guaranteed to be contiguous.

Test: Booted the system, verified devices were published.
Change-Id: Ic573c10adf847d34c7b8aeac788082322d80adde

Test: fvm-test
Change-Id: Ie24d7de8c7fc2208f573f1055a6f5f13fb0a795d

The formatter was unable to track the beginning of a method declaration
if there was no method ordinal present.  It therefore did not know how
to indent the next line.

FIDL-395 #done

Test: Added a test to badformat.fidl.noformat
Change-Id: I68a077ed15bc4179a65a2c7ccf672ef4767aee47

TESTED=`ps --help`

Change-Id: I51e4a63487527595dd4746a7dc10d55606b92fa7

Adds a "pave" command which will unpack a sparse
file into a full-fledged FVM image.

Test: TestPave in fvm-test

Change-Id: If87342aeed891a52b7bf5084485935a16c186ce0

This adds a new mode for kcounters that treats the value as a maximum
across cores, rather than a sum. These are declared with KCOUNTER_MAX()
and modified with kcounter_max().

Additionally, adds two new max counters, one for live handles, and one
for maximum allocation size in cmpct_alloc().

In implementing this I noticed that I think `kernel.handles.live`
probably doesn't do quite what it wishes it was doing (maybe?) and filed
ZX-3015.

ZX-3092 #comment [counters] Support a "max" mode for kcounters
ZX-3105 #comment [counters] Support a "max" mode for kcounters

Test: CQ, manual inspection of `k counters view`
Change-Id: I556fe74dfcb54e1fd61421fb13b6be65d4687243

Test: Ran /boot/test/ddk/driver-tests
Change-Id: If3b8578d545ce80450f5d4adc622d7e912dc672e

Adds the ability to calculate the unpacked size of a
sparse or fvm container on disk, as well as to check
if a sparse file will fit within a given disk size.

ZX-2704 #done
ZX-2981 #done

Test: TestDiskSizeCalculation in fvm-test

Change-Id: Ibc68780f22bb358ce2847c8df284d5f4ded3ae9e

Test: No behavior change
Change-Id: I3cc379cc6bb1f4dedfafc66bb4563c574da1a485

Test: display-test, set_root_view image_grid_cpp
Change-Id: Iaeb9e0f3d3bb72424fb06ec45f99094869bdbbf6

Test: Ran formatter, verified output.
Change-Id: Ifaa8a0b12825153cadea19c6f11068afa09bf2b4

The only caller of Lookup which passed pf_flags requested just a single
page. Replace that call with a call to GetPage, and then simplify
Lookup so it won't have to deal with delayed page requests.

Test: runtests, boot zircon_guest
Change-Id: Ic92f833d010d4555cbc4677e6df7c42505de2521

Also add missing size assertions to port_dispatcher and update the
assertions to check against a constant size.

Test: runtests
Change-Id: I9531a3e4b55dc24badde13e758f306e322ae5cb6

With usb request having private regions for each layer in the stack,
the complete callback and their contexts should not be part of the public
part of the usb-request. This changeset modifies usb protocol so that
the complete_cb and its context are passed in as parameters.

TEST: USB peripheral role functions work on hikey960, USB ethernet on mt8167s

Change-Id: I144d4ee98a96077fe163536628ceb9e7850656d8

TEST: No behavior change
Change-Id: I718a814d032df583c6bc791b948c97a27267ab7b

This allows passing the TEST_LOOP_RANDOM_SEED through other means if the
environment is not usable.

Test: async-testutils-test
Change-Id: I51396388a232546dd1cdea583df2f944390af870

Test:

Change-Id: I457d6e14f8de96bf938a2c12297d0bdecaf3e4c8

TEST: No behavior change
Change-Id: I1529495bd539647e22988be1098cb704dc5acbef

TEST: No behavior change
Change-Id: Ie9fb329ef6131f4fc799345540cae322dd4e1c33

TEST: No behavior change
Change-Id: I6f3089887d3ab31dcf932832d5f1e58be54d2afc

This reverts commit 9b5ac003.

Reason for revert: This breaks global integration, due to the Broadcom drivers in Garnet.

../../garnet/drivers/wlan/third_party/broadcom/brcmfmac/usb.c:271:44: error: too few arguments to function call, expected 4, have 2
    usb_request_queue(usb_proto, urb->zxurb);
    ~~~~~~~~~~~~~~~~~                      ^
../../zircon/system/ulib/ddk/include/ddk/protocol/usb.h:191:1: note: 'usb_request_queue' declared here
static inline void usb_request_queue(const usb_protocol_t* usb, usb_request_t* usb_request,
^
1 error generated.

Original change's description:
> [usb proto]Add complete cb as a parameter
> 
> With usb request having private regions for each layer in the stack,
> the complete callback and their contexts should not be part of the public
> part of the usb-request. This changeset modifies usb protocol so that
> the complete_cb and its context are passed in as parameters.
> 
> Test: zircon build. Will be changing garnet also and then will do basic
> sanity test.
> 
> Change-Id: I42ab688b3db2aaca56f05d67a69d94b61199c256

TBR=voydanoff@google.com,jocelyndang@google.com,ravoorir@google.com,surajmalhotra@google.com

# Not skipping CQ checks because original CL landed > 1 day ago.

Change-Id: I886923ac032e3713a9426b992a2dd542357c25c9

We may have as few as 16 priority levels, so we should appropriately set
the priority level.

MAC-180

Test: Ran "zircon_guest" on a VIM2.
Change-Id: Iaf30b9d0eeb8a9a9b1dda1e16acc93b806ba1b8b

This reverts commit bcc6f70f.

Reason for revert: This is causing NetStreamTest.BlockingAcceptWrite to timeout.

Original change's description:
> [fdio] do not accept sockets with a full fd table
> 
> If the fd table is full, pulling sockets from the control channel results in
> consuming the incoming connection, but being unable to actually service it -
> we then immediately close (which then ends up in linger). Other processes
> would then not be able to pull these sockets (in a load balance scenario),
> and netstack also potentially loses the ability to apply policy from having a
> full listen queue. Instead, try to allocate an fd up front, and only if found
> accept the socket from the control channel. This is more conformant with
> common accept behaviors.
> 
> Test: install a package with 5500 blobs with a package server routed via a remote port forward over ssh. Observe successful install rather than failure.
> Bug: PKG-371 #comment fdio: do not accept sockets with a full fd table
> Change-Id: Idb5513ada590b0d8da2f5cd68b4d734765e438d4

TBR=jamesr@google.com,mcgrathr@google.com,raggi@google.com,marshallk@google.com

# Not skipping CQ checks because original CL landed > 1 day ago.

Bug: PKG-371 #comment fdio: do not accept sockets with a full fd table
Change-Id: I31716bafa4cee2116cca16541d4348025c571e56

Only C++ const-correctness is important, not whether the underlying
kernel object is mutated.

Test: Ran system/utest/core/interrupt.
Change-Id: I72529d61063ce2c360e6b13c429902a521e2133a

Since we only care about C++ const-correctness, and not whether the
operation is mutating the kernel object, we should mark these methods as
const.

Test: Ran system/utest/hypervisor.
Change-Id: Ic10e9b86a24dd7a91a49d88efc2c4f98a78ce756

With usb request having private regions for each layer in the stack,
the complete callback and their contexts should not be part of the public
part of the usb-request. This changeset modifies usb protocol so that
the complete_cb and its context are passed in as parameters.

Test: zircon build. Will be changing garnet also and then will do basic
sanity test.

Change-Id: I42ab688b3db2aaca56f05d67a69d94b61199c256

- Drastically simplify lifetime management of the FIFO
server, removing a "lock, completion signal, threadcount,
and boolean" with a ServerManager class that provides
a simple API.
- ServerManager utilizes a thrd_t and an atomic to identify
if a FIFO server is running, and if so, in what state. The
most notable improvement comes from ServerManager's ownership
of the background thread, rather than detaching it after
launching.
- Isolating this logic into a "ServerManager" class will
make this API trivial to unit-test, as soon as the "blockserver_*"
APIs are abstracted.

Test: /boot/test/sys/ramdisk-test

Change-Id: I07c951db31c663b0aab68da2794b77408723ef07

This enables:
- RAII wrappers around handles, memory.
- DDKTL usage, to avoid casting between "void*".
- Automatically initialized mutexes.
- Lock thread-annotations for safety.

In the future, this will also enable improved encapsulation of the
BlockDevice object, but structural changes have been minimized for
the current patch.

Test: /boot/test/sys/ramdisk-test

Change-Id: Id041bd4cc5f541620e4829427fac3c964700b13b

GetStats is handled by the block core driver, not implementation
drivers.

Test: CQ

Change-Id: Iec7f3cd8c00af4c175b7f0b72610931ffcd91f4e

This makes the diff from C --> C++ easier to see.
The current block.cpp is an exact copy.

Test: None (dependency not added to build)

Change-Id: Ifec4f559f0f012f094784d61bbe6f836d80ddf34

This is not applied to all syscalls yet, as there are many where the
argument name in abigen doesn't match that in existing documentation.

So for now, use a whitelist to only update ones where the argument names
match. This fixes a few typos and missing 'const's but otherwise doesn't
change the SYNOPSIS section for whitelisted syscalls.

ZX-968 #comment [rights][docs] Generate some SYNOPSIS sections from abigen
ZX-2399 #comment [rights][docs] Generate some SYNOPSIS sections from abigen
ZX-3106 #comment [rights][docs] Generate some SYNOPSIS sections from abigen

Test: CQ
Change-Id: I74acd98a13ad5b78bf2499657824cda35d0b6a4e

TEST: No behavior change
Change-Id: I6b907e1d36432ae5fdea422e81faaa727e062f9a

If the fd table is full, pulling sockets from the control channel results in
consuming the incoming connection, but being unable to actually service it -
we then immediately close (which then ends up in linger). Other processes
would then not be able to pull these sockets (in a load balance scenario),
and netstack also potentially loses the ability to apply policy from having a
full listen queue. Instead, try to allocate an fd up front, and only if found
accept the socket from the control channel. This is more conformant with
common accept behaviors.

Test: install a package with 5500 blobs with a package server routed via a remote port forward over ssh. Observe successful install rather than failure.
Bug: PKG-371 #comment fdio: do not accept sockets with a full fd table
Change-Id: Idb5513ada590b0d8da2f5cd68b4d734765e438d4

TEST: No behavior change
Change-Id: Ie4e1292a13d095e534bc43b3f0fa2ad9f9501367

This exposes the zx_interrupt_bind_vcpu syscall through the
zx::interrupt object.

MAC-180

Test: N/A
Change-Id: I3353d6546f6471aee179b74086343f570bcedf59

Also leave a note to discuss the many other port packet types in a
subsequent CL.

Test: no functional change
Change-Id: Ie242767af1d7f37ce3db099a73fe9587fd41eab3

Fixes a few small inconsistencies. (zx_ vs. not prefix, '## NAME'
missing in few files, missing spaces.)

I didn't take a position (yet) on "lower case no period" vs. "Proper
sentence with trailing period.", but my left eye was twitching.

ZX-968 #comment [rights][docs] Move NAME short description to abigen
ZX-2399 #comment [rights][docs] Move NAME short description to abigen
ZX-3106 #comment [rights][docs] Move NAME short description to abigen

Test: CQ, no intended functionality change
Change-Id: I3a74062e7c599701ca5608731eb3fc27446ad044

TEST: No behavior change
Change-Id: I3264c2014e2a27fc0a14f11885428269479404bb

ZX-968 #comment [rights][docs] Handle arrays and structs (just barely) in doc gen
ZX-2399 #comment [rights][docs] Handle arrays and structs (just barely) in doc gen

Test: CQ
Change-Id: I8e63044830b48c315e3fc6c8dd766609fc2c161f

Use the zx::port object, rather than the raw handle, when binding the
interrupt.

MAC-180

Test: Booted on a VIM2, but don't have a mt8167 to test with.
Change-Id: Idd9e8cc20359fcccb792b71a9a68ef267df15d92