We no longer need this header.

Test: No behavior change
Change-Id: I03679671db7c478709c0f78f815c310cb238b3c5

Test: build/ran kpci with graphics under qemu

Change-Id: I87a256ccc299f11cd1fd4aa4667a9acf89d364b7

This CL removes declarations from remoteio.h that are redundant with
declarations in io.fidl.

Test: No behavior change
Change-Id: Id23f270b69b575b104bf00c2bed0140c6408fff0

XRay needs to retrieve the innermost VMAR to allow changing its
permissions when patching function entry and exit points. This is
currently not possible since the dynamic linker always closes the VMAR
handle. To support XRay, when the dynamic linker is being built for
XRay, we don't close the VMAR handle and provide a new runtime function
__sanitizer_change_code_protection which allows caller to change the
protection of its own code segment.

Test: manual
Change-Id: If6a780d29cce389068574cc8b50d0f2387f8e3a3

TEST: install-device-partitioner-test
Change-Id: I8be7223ca9d3721c93728552290d5e16cf41cc24

This will allow the user to receive fewer callbacks for
isochronous transfers.

Implementation will be in next CL.

Also removes old cb_on_error_only implementation, which had
issues detailed in ZX-932.

An example of how the user would call the API:
https://fuchsia-review.googlesource.com/c/garnet/+/227672

ZX-932 #comment

TEST= set_root_view video_display, does not affect current video implementation

Change-Id: Ia06c8a6ebccd1042e667a9e2d623c3b57265a1ff

test: dictionary
Change-Id: I92f2bea38b9897f91c20b07ed6f535a02ed2b7ec

In an error handling case, zx_task_resume() was still being used. It's
deprecated and to-be-removed shortly, so switch to the newer function.

ZX-2720 #comment

Test: CQ
Change-Id: I49d2b0050bb8edec93fe9de644c157e294d0c4b2

Paring down the number of references to this so that it can actually be
removed.

ZX-2720 #comment

Test: CQ
Change-Id: Ibb0d91809f5914bde27a0bef3a76022ed31ac6a1

DX-636 #comment

TESTED=`crasher` on x64 and arm64

Change-Id: I5f9a9ace2f2cf066e50c5d3960e597658508651e

The only non-test user of either count was just using the commit count
to verify that the whole range was committed, which is redundant with
the commit being successful.

Test: k ut, core-tests
Change-Id: I51ba403c5f1711c99108a538682b098bb6f140fd

This library was based off of the code that was previously being used.

Test: Booted with this change
Change-Id: I2115e2f6350d59e34e1b5f361c4717f597abf938

This is an implmentation of FTP-020.  The next step will be to deprecate and remove
ordinals from existing FIDL files.

To make removal easier, this CL introduces a --remove-ordinals flag to fidl-format.

Test: Unittests added for compiler and formatter.
Change-Id: Id5c10f84bde95656fca3368a4fb21ca00310197b

This code will soon be shared by the new bootsvc and by fshost.

ZX-3027 #comment Lift bootfs parser into shared code

Test: Added test suite.  ZX-3034 is a bug for making even more thorough
tests.

Change-Id: I583130ced2675d1471d86e3a51ec93c7f28f5bfe

This makes more syscall pages link to `object_wait_async`
when they already link to the other wait calls.

TEST: No behavior change
Change-Id: I8770a9baf9ab5aea0e00719f9f7f262994ddc795

NET-1915 #comment

Test=n/a

Change-Id: Id1c5a9bb56059ce92880df1b9789c1d45b437daf

there doesn't seem to be a self_dump_func or an actual use of
crashed_thread

TESTED=`crasher` on x64 and arm64

Change-Id: Ib5c676fea8fc40679a894d4665b29490a0784497

Currently, there is only 1 list_node_t pointer in usb_request_t and
differnt drivers in the usb stack that track a usb request enqueue it
using this node. This will lead to bugs where the request is simultaneously
included in multiple lists changing the next and previous pointers of the node.
This changeset makes use of the xhci's private list_node_t pointer
that is only used in xhci driver. In xhci, a particular request is first
queued in queued_reqs, then moved to pending_reqs. At a particular
time,it is part of only one list.

Test:
vim2: -fx serve; plugged in a mass storage device.
      - lsusb; iochk
pixelbook: -fx serve using usb ethernet adapter
           -plug in a usb mass storage device.
           -lsusb; iochk.
           -audio info using a usb audio headset.
astro: - fx netboot; lsusb; Made sure the usb ethernet adapter showed up.

Change-Id: I172e36046366b41d8e9cc1bd5a50ba4264618475

This CL adds basic fuzzers for the lz4 library.

There are two fuzzers being added. Once is a "roundtrip" fuzzer, that
will try to compress the input data, decompress the result, and check
that they are equal. The other fuzzer just tries to decompress random
data and crashes right away (until we fix it).

TEST=`fx fuzz start lz4.{fuzzname}` in a pixelbook running fuchsia.

Change-Id: Ic3b9ab98e3c8cb1d345605ad3b0b667dac8b8c5f

This adds the framework for a userspace PCI driver using
DDKTL. PciConfig has been ported from the kernel bus driver
and proxying config access to the pciroot protocol has been
implemented.

Test: Confirmed the output of the root 00:00.0 device config
space dump, and ensured kernel pci behavior remains unchanged.

Change-Id: Ic7ec3dc72ce2ff34196ff624234127937383459c

This CL adds a zxio backend for vmofile but does not migrate the
production codepath to zxio.

Test: vmofile-test.cpp
Change-Id: I84721065540d304d257269a8527eb1e7a7926182

Currently in sdmmc_bind function, we always create a block device. Later
we probe the device and once we know its a sdio device we remove the
block device and create a sdio device and we end up calling
device_remove function twice. This changeset creates a dummy device and
then creates child based on the result of the probe. There are other
use after frees that are also fixed with this changeset.

Test:
Vim2: With ASAN: Booted with ASAN. Without this change, sdmmc crashes
in sdmmc_worker_thread. With this change, no sdmmc crash. both
SDIO/sdmmc devices show up.
Without ASAN :
- fx serve
- iochk on sdmmc device.
- wlan scan/connect/ping.
- run wlan-smoke-test GoogleGuest.
Astro: Without ASAN:Verified wlan scan/connect.
With ASAN: No sdmmc crash in sdmmc_worker_thread like on TOT. Now that's
resolved and we proceed further loading brcmfmac driver. However there
is another crash in aml_sdmmc_irq_thread. I will raise a separate bug and
submit another changeset for it.
Pixelbook:
- fx serve.
- iochk on sdmmc device.

Testing on pixelbook is in progress.

ZX-2951 #Done.

Change-Id: I9694b82138d9ba98ee66ca2da429a048479771db

Other drivers can use this code for unit tests.

Test: current (ram-nand-test)
Change-Id: I2c2480ba236806f5f3ad7d300565996ee786288c

There is nothing left in this file and we can remove it.

Test: No behavior change.
Change-Id: Ie3fbcf522980bf7c3f7b3a281f4c5a867e834a34

This code is used for file servers not file clients. Rather than linking
the code into the client library, fdio, we now link the code into the
server library, fs.

Test: No behavior change.
Change-Id: I62717d718ead798cafecc87e1d5924d457771aac

This reverts commit fb223fcb.

Fixes the previous breakage by adding padding to
fuchsia.net.socket.AddrInfo.

Also fixed FIDL C bindings' encoding of optional strings.

This is part of a hard transition.

Tested: /pkgfs/packages/netstack_manual_tests/0/bin/getaddrinfo_test -n
google.com

Change-Id: I147ddeb55f8779f94e5b937508e502d8f23863ce

TEST: No behavior change
Change-Id: I6c212b9903408ae3488659ef835555ecead4b22c

TEST: No behavior change
Change-Id: I59b23d275e6a33d8585c211eadf5d60c1ed8cbe6

TEST: No behavior change
Change-Id: I3dc1718bb3bc313d871cea6c1b14b4137a9819e5

by splitting process_report into print_debug_info and
print_debug_info_and_resume_thread, we allow dumping the debug info
without resuming the thread.

while crashanalyzer.h is not visible from other repos, this is just the
first step. we will move the helper in another patch.

DX-636 #comment
DX-653 #comment

TESTED=`crasher` on x64 and arm64

Change-Id: Ic90035f012feb90eec3fba8aa0449301ac8ae295

This reverts commit f2cc5673.

Reason for revert: Interrupts are not masked when we read IAR, therefore we do need a DSB.

Original change's description:
> [gic][v3] ICC_IAR1_EL1 is self-synchronising
> 
> In the GICv3 manual, section 8.2.14, it describes reads of ICC_IAR1_EL1
> as self-synchronising, therefore we do not need the DSB in
> gic_read_iar().
> 
> Test: CQ, as I'm unable to test this on hardware.
> Change-Id: Ieb0b516a19c3bc4db0be2b5215faebac085479aa

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

Change-Id: I2d404cb5f29e339c48125721ea3f1d198f4b2246
No-Presubmit: true
No-Tree-Checks: true
No-Try: true

This is a reland of 10fd5d15

Change-Id: Id1ed9142fe098c25991b91286212d99f513f18a8
Tested: runtests -t nandpart-broker-test

Test: Function is unused, but confirmed the build is successful
for both upci and kpci.

Change-Id: I2b1e992fab43cded8e345f448573240f20f92de7

list_node_t should be private to each driver, so that it doesn't
accidently get double used in multiple layers of the driver stack. This
is following the same design nand and block protocols currently use, and
the one the usb protocols are transitioning to. The EthmacNetbuf struct
defines the public portion of the buffer, and a each driver in the stack
may define a private section afterwards. They report their size to
higher level stack via the Query call. The highest driver in the stack
(ethernet), is responsible for allocating a buffer at least as large as
it's device's parent.

Tested: Paved pixelbook, vim2, and astro
Change-Id: Ie066adbdf112e58d10ad114ce8fbdd78edecb513

Tested: Compiles
Change-Id: Id97cf8d5428a63d576ba39d1cb9464ba66e78e31

TEST= usb-test-fwloader

runtests -t usb-test

Change-Id: I4fa490f6837d533d9e90dee70fe35588b239d3ca

Test: built system, ported some garnet fidl C++ tests to linux

Change-Id: I21df2e690760709c96d69576377db7ff146ef415

This rolls the symbolizer that has ids.txt rel paths

Test: By hand.
Bug: TC-290 #comment roll symbolizer into zircon
Change-Id: I25e3648656a35f8637b59016b2bf299836ff0ad6

Tested: Compiles
Change-Id: I71805d951e28bb9c00fe2339a8eba99581e50bf1

Tested: Compiles
Change-Id: I959bfd02e14e8e0c0be51e69d7c2377c8d8e112c