Although there is still heavy reliance on ordinals as constants
outside generated FIDL code, this patch removes the filesystem
dependence on relative ordering of FIDL ordinals by instead
utilizing the "try_dispatch" family of functions.

Rather than routing to the appropriate dispatch function, the FIDL
handler will now "try_dispatch" to all possible interfaces,
checking against ERR_NOT_SUPPORTED until something else is returned
or the list is exhausted.

This is less efficient, but acts as a workaround for FTP-20.

ZX-3063 #done
Test: CQ

Change-Id: Iff1736757417e2b44765ae0083fc6d16022102fe

Given previous changes, enum member values are now validated and
resolved at compile time and should no longer need to be parsed by the
C backend. This change removes the parsing from the C backend and
instead, reads the resolved values.

FIDL-304 #comment

Test: No regressions on C bindings.
Change-Id: If9e97afc97da8b1c3f66e5cd337d2b4c3ef37c94

TESTED=cq

Change-Id: I96797b05b320d86f8ac12e53d89ce6198bb10b3d

Instrumenting raw to flat conversion to check attribute placement, and
adding a central place where placements can be checked. Using this to
prevent attributes like [Layout] or [FragileBase] to be incorrectly
placed.

Test: make -j 72 HOST_USE_ASAN=true tools && ./build-x64/host_tests/fidl-compiler-test
Change-Id: I21cdce25b0c06f110938a3baeff7ca4fd8dccfd4

See https://fuchsia-review.googlesource.com/c/zircon/+/214943

FIDL-204 #comment

Test: cq
Change-Id: Ie831e81a253331cd3769728d2e121feb6d96f45d

TEST: CQ

Change-Id: I2a160577bbf1d7ed503f2aee0c0b0a263b4e18a7

Test: no functional change

Change-Id: I7701e63d509a3c823fab0af28dd0f7443c6aa41b

ZX-2997 #comment In Progress
Test: /boot/test/fs/blobfs-unit-tests

Change-Id: I1d0d6d618cc248eb82dd12caf44ccc850e532852

Currently, all audio drivers reside in zircon/system/dev/audio.
Certain new audio driver work might benefit from being in garnet
instead. Doing this requires these two zircon-based sourcesets to be
exported to zircon/public/lib.

Test: build (with custom garnet project that consumes it), CQ

Change-Id: Ic194a46d56f9a5cde5da890c59b39d780c6ca152

Previously, members of enums were not validated and would propagate
errors into the backend. This change moves the validation into the
frontend. Validation includes:
  (1) Member value resolution
  (2) Member typechecking
  (3) Member name uniqueness check
  (4) Member value uniqueness checks

FIDL-371 #done
FIDL-374 #done

This change also includes new tests for validation.

Test: Test fidlc on both valid and invalid enums and run new test cases.
Change-Id: I16b70bda994f0b778af7232513c3beac2691a829

Peripheral role only. Support for host role comming later.
Uses fifos for IO. DMA support not yet implemented.

TEST: netboot zircon over CDC ethernet function

Change-Id: Id52ee06a138412837893e15e2862b82d9b3afaa4

Adds RAII interfaces to allocate nodes and extents.

Additionally, defines an Extent structure which will (in later patches)
be added to the blobfs inode.

ZX-2997 #comment In Progress
Test: /boot/test/fs/blobfs-unit-tests

Change-Id: Ia05e0c60c61b679319ea23cf29915b0f4bd00f53

Refactor thread_sleep_etc to accept slack parameters and the current
time. In a future CL, (most?) calls to thread_sleep_interruptable will
be replaced with calls to thread_sleep_etc that pass slack arguments
obtained from the ThreadDispatcher.

Bug: ZX-931 #comment Refactor thread_sleep_etc
Test: CQ and "k sleep_tests"
Change-Id: I387aef7e7af9112102b947b1bf654bf4b66897db

With the recent changes to usb stack, each driver in the usb stack has its own list_node_t*
so as to avoid stamping on other driver's lists. This changeset adds a
usb_request_internal_t and some helper functions for convenience of all
the clients.

Test: Build. Tested that the apis can be used in a client

Change-Id: I8494060293b1c3c6e4c773f7db39c50af124bfd7

Test: n/a, docs

Change-Id: Idff4e373724539edf445f196a4b21a129c77cf58

ZX-2967 #comment [rights] Make svchost job copy use MANAGE instead of READ|WRITE

Test: boot garnet
Change-Id: Iec71361ebdf52744e179d2cc357f3bdfba90fcd3

Propagate AddPage failures due to the page already existing. Existing
callers all know that a collision cannot happen, as they either use a
new page list or they query the list first.

Test: k ut, core-test
Change-Id: I7ba60889c3bfea6ea9e3847d9bfafd5e9ba2d2a7

This reverts commit c8d3e357.

Reason for revert: mt8167s no longer boots with this change.

Original change's description:
> [arm][gic] Change EOImode to separate deactivation
> 
> Set the EOImode so that EOIR only drops the priority, and DIR is
> required to deactivate the interrupt. This allows the hypervisor to
> correctly take control of physical interrupts, without affecting
> operation of the host.
> 
> MAC-180
> 
> Test: Ran on a VIM2 (for GICv2) and QEMU (for GICv3).
> Change-Id: I110eda0b609ff7acca97f4d3baad5cbeab1b3b9c

TBR=travisg@google.com,abdulla@google.com,maniscalco@google.com

Change-Id: I6b7b4ada61ab4ec1f8f9949aa3af6ac53892626b
No-Presubmit: true
No-Tree-Checks: true
No-Try: true

Bug FIDL-204 is resolved, no need for manual alignement for data used in
C bindings.

Test: fx run-test netstack_getaddrinfo_test
Change-Id: I755f3412463a6e7c89b1fac1706259b6f126d8e2

Clients have already migrated to using rx_buf_available from the
ZX_INFO_SOCKET topic for zx_object_get_info.

ZX-2798 #fixed

Test: socket_zero_size
Change-Id: If1647dc6550f0b3011afe66dd892d79e4876b45d

Test: no functional change
Change-Id: I9a39cd66a1bf4e26d68d4dc7c54f228075c12081

Tested: CQ
Change-Id: I226e1cf1b17453e5cc066c989773e1e1ae4db8fc

Having it unbounded was resulting in overflowing the uint16_t request blockcount

TEST: paving on mt8167s_ref board
Change-Id: I03910a1ce2fa4512e80a2a785bd0b9aacce3d4b6

This is for use when writing one's own simplifying wrapper macros.

Tested: CQ
Change-Id: Id6a0f14e0c29fb396047b2800e08435c84df07c8

Set the EOImode so that EOIR only drops the priority, and DIR is
required to deactivate the interrupt. This allows the hypervisor to
correctly take control of physical interrupts, without affecting
operation of the host.

MAC-180

Test: Ran on a VIM2 (for GICv2) and QEMU (for GICv3).
Change-Id: I110eda0b609ff7acca97f4d3baad5cbeab1b3b9c

Note that a virtio flush request refers only to flushing volatile writes to
persistent storage wheres a flush operation in the block protocol is
also a barrier. Since the virtio specification allows devices to service
requests in any order, we must wait for pending transactions to complete
before and after making a flush request.

ZX-2724 #done

TEST=guest_integration_tests + fx run -V

Change-Id: I201688f8a468aca02d13c06948e25c52772ef4ce

more specifically:
xprintf -> FS_TRACE_DEBUG
fprintf -> FS_TRACE_ERROR

ZX-3057 #comment In Progress
Test: CQ (no visible runtime changes)

Change-Id: I0e5900ea4a580e9e04ad1982552d5399da1c2a6c

This macro takes a hidden parameter of the trace context whereas all the
other TA_*() macros don't. The presence of this hidden parameter makes
extending the API problematic, e.g., to introduce TRACE_DECLARE_ARGS().
TA_STRING_LITERAL is currently only used in one place:
garnet/lib/machina/virtio_net_legacy.cc with very short strings ("RX", "TX").

It can be added back later after we think about how we want to handle it.

PT-66 #comment patch

Tested: CQ
Change-Id: I92b754e53face397c59357bd211b4d612132380e

Test: Tested on both Astro and Sherlock

Change-Id: I3857146c98349b601147f4c4bc5654f17bd51b54

... to helper functions. This reduces the amount of code generated
at the TRACE_*() call site.

trace-benchmark results show modest improvements for large numbers
of arguments.

E.g.,
* oneshot, 16MB buffer: TRACE_DURATION_BEGIN macro with 0 arguments
Before: total (usec): min: 10579.976, max: 11130.338, ave: 10644.971
After:  total (usec): min: 10594.704, max: 11167.938, ave: 10719.786

* oneshot, 16MB buffer: TRACE_DURATION_BEGIN macro with 1 int32 argument
Before: total (usec): min: 13495.626, max: 13869.269, ave: 13538.099
After:  total (usec): min: 12905.281, max: 13237.774, ave: 12945.262

* oneshot, 16MB buffer: TRACE_DURATION_BEGIN macro with 8 int32 arguments
Before: total (usec): min: 34639.172, max: 34685.566, ave: 34653.770
After:  total (usec): min: 31806.008, max: 31974.891, ave: 31855.652

Size reduction: utest/trace/trace.elf.strip went from 510000 bytes to 456752.
Analysis of utest/trace/event_test.cpp.o:test_counter shows a 53 byte code
reduction for this trace invocation:
    TRACE_COUNTER("+enabled", "name", 1u, I32_ARGS1);
and a 259 byte code reduction for this trace invocation:
    TRACE_COUNTER("+enabled", "name", 1u, I32_ARGS4);

PT-65 #done

Tested: CQ
Change-Id: Ia57bc64f1b9c2df2263709449b155cfcb1bab104

FIDL-204 #done

Test: make -j72 && ./scripts/run-zircon -a x64 -c zircon.autorun.boot=/boot/test/sys/fidl-simple-test
Change-Id: I01619cb42f165acbb2e4947e9e6d48f207773f1a

Change MMPLL from 380 MHz to 400 MHz so the bus
clock can run at the highest frequency allowed by
the spec. The eMMC controller does not appear to
support HS400 mode, so this is the best data rate
we'll get.

ZX-2783 #done
Test: slight performance improvement confirmed
      with iotime.

Change-Id: I2607d338647ed75798f07915386965a58c655d4c

This seems to work (all of devmgr's launches succeed), but I would not
be surprised if "something" breaks.

ZX-2967 #comment [rights] Have devmgr_launch() use MANAGE rights instead of READ|WRITE

Test: CQ, manual poking around
Change-Id: I866d1a2c783b82d88c52c59ddd710f189a1efeeb

Test: clang lock annotation checking
Change-Id: I4f708e1449db34f0694ce1635eecd8d272f460f6

Test: no functional change
Change-Id: Id68768b6e3ca29a3a58c95737cc9b14cd468e0c6

Test: no functional change
Change-Id: I17dc9b7cfc9ddd6eba80165e777310ce2639a334

Add "DEBUG" and "INFO" levels to FS_TRACE macros. Currently compiled
to no-ops; these will eventually invoke the unified logging API for
Zircon.

ZX-3060 #comment In Progress
Test: CQ (No runtime visible changes)

Change-Id: I23c8877f9b0478766b77f3e2ad8be8e827a75b65

ZX-3012 #done

Test: hid uapp in sherlock.
Change-Id: I4ddec9b597e1dacd48abe98ba8971a736d2e9501

Edge interrupts do not require the state machine handling required by
level interrupts in the kernel and work as well in hid-buttons.
Issues with level interrupt handling has been seen at least on one device
that is being tracked by ZX-3077.  Since we have support for edge interrupts
in all HW that uses buttons today, switched to edge.

Test: hid uapp in astro and mt8167s.

Change-Id: I36725ee87baa7b007b1c523c17d2daca5efda65b

Add PROCESS management to jobs by default, add THREAD management to
processes by default. The logic here is that job handles need to be able
to create processes inside themselves, and that process need to be able
to create threads inside themselves.

Additionally, accept MANAGE_PROCESS on a job as sufficient to
process_create() with a fallback to WRITE for all the exisiting callers
that are calling with WRITE.

ZX-2967 #comment [rights] Add MANAGE_PROCESS to DEFAULT_JOB, MANAGE_THREAD to DEFAULT_PROCESS

Test: CQ including garnet and topaz
Change-Id: Icf767a369299596c2d5974225fb9e31fa59a923b