Change-Id: I1f976e1602f5c9215a1bf915013132bcd9698c5b

Change-Id: I0149c24b3b9f0457ed4e2de4ced0faeb9c5a94c6

Step 1/5

NET-9

Change-Id: I7ff2f20b1a1c743f1a33ae5b6b22119656173b07

Step 3/5

NET-9

Change-Id: Icfc0d2ae2b1bc237f136cb9ccacea83bf39bfe1a

Step 5/5

NET-9

Change-Id: Ib2dc108feb26045f6f8e33996d666fc756648805

Change-Id: Ic5b672665e2695feade1f05871794a7c06e87e77

Change-Id: I05877cc081d03752335c8d078f59760b9efaa095

- only supports devices with a single namespace
  (so far every m.2 ssd I have seen and Qemu comply with this)
- does not support SYNC operations yet
- does not do proper cache and barrier ops for arm64 yet

Change-Id: I06c6674695ea3222d69fa244d82591c463a0fec2

Change-Id: I49ecd0797938b3975701ccea566d90023d7d8928

Change-Id: I8c08878e3609114e4bfd1027803ad475b4a68b31

For block requests which are larger than the maximum
transfer size, don't wait for each sub-chunk to complete
sequentially. Instead, IotxnQueue all sub-chunks at once,
and use a simple counter to track when they complete.

Change-Id: Ia074e9f32d34fae870ba85a5e2902ac6c2858937

Change-Id: Ic914253f30047db4b7d274946b05dd585e92a80c

Change-Id: Idbc2721499a5b18acbb395d26b63bf13d08c79a0

get_dma_offset() was a hack added to support RPI3. Remove it now
that it's no longer a target.

Change-Id: Ib79bfd330a42a6e51ceea3cb9c7d828ecf691825

driver, and related struct definitions.
This borrows heavily from the audio driver.

Actual implementation will be in next CL.

ZX-1129 #comment

Change-Id: I18e84b4c91d7b67c5a5f7017f5fd86fb31216a7f

…EDULE

Fortunately, true and INT_RESCHEDULE both have the numeric value 1, so
this should make no difference.

ZX-1490

Change-Id: Ib121626fa3f94973b1f77913d432eadc77497786

Example usage of async::Wait:

    class Foo {
        Foo() { wait_.set_handler(fbl::BindMember(this, &Foo::Handle)); }
        async_wait_result_t Handle(...) { ... };
        async::Wait wait_;
    };

With async::WaitMethod, that becomes:

    class Foo {
        async_wait_result_t Handle(...) { ... };
        async::WaitMethod<Foo, &Foo::Handle> wait_{this};
    };

When async::Wait is used with fbl::BindMember() like that, a single
callback from the event loop to Foo::Handle() will involve 3 indirect
jumps plus 1 conditional jump (which generates a lot of wrapper code):

 * An indirect jump to async::Wait::CallHandler().
 * An indirect jump to fbl::function's virtual operator().
 * A conditional jump to check whether &Foo::Handle is a virtual or
   non-virtual method.  (This is part of the semantics for C++ method
   pointers.)
 * An indirect jump to the Foo::Handle().

In contrast, async::WaitMethod<> will do only a single indirect jump,
and no conditional jumps.  (The conditional jump is avoided because
the callback method &Foo::Handle is passed as a template parameter and
it is known statically to be non-virtual.)

Change-Id: Id581b39eb9939d97148529e5f3453243c72bafc0

The magma implementation of acquire_or_release_display makes a call to
the ownership callback, which takes a lock. Change 866e4bc made sure
that the framebuffer's parent device is never used after unbinding, and
in doing so it moved a call to mtx_lock before calls to FB_ACQUIRE and
FB_RELEASE, resulting in a double locking issue. Since the magma driver
never unbinds, it isn't necessary to move the lock for that driver, and
the framebuffer driver will go away before more GPU drivers are added.

Change-Id: I1790ec476420640a7453fe6bcca0f22a9932eb40

Ran clang-format on the dpc files as well

Change-Id: Id333fa05b63f1758dd505cb3fd7b6d397d3b247c

Change-Id: I2e83a566d78d8a57f9e1ac3bde892dc043ae78b4

…onsistency

ZX-1490

Change-Id: I653b55552c901ad5ee43fb3ba3826b51a01038d1

This function is not used outside its file.

Change-Id: I71350f2e3c347003bb0c94e7a921f7832a703f9b

Accepting all x86 image files and ignoring a subset based on system
detection is a version worthy change.

Change-Id: Ibec8bbab761e4c4c5647a8b9d9be905c70ff5f73

This should be used as a static library within Zircon and a source-only
library outside of it.

Do mark it for source packaging for import int Garnet and above

Change-Id: If1c7b44aaaba4b1adf0dd5ba0960c7d3e3b0ded9

Change-Id: Ie0c48f6cd559dc6587e181f940d1a5b7675fb9d8

The plan here is that we will use a protocol more tightly aligned
with the needs of the block IO layer (including constraints that
make things much more consistent and less open ended than the
more free-form iotxns did).

The expectation is that the only client of this protocol will be
the generic block mid-layer driver.

The deployment plan is:
1. Agree on the new protocol
2. block mid-layer driver will support new protocol, if present
3. block drivers will support new protocol
4. once all block drivers support new protocol, we retire iotxn_queue()

Change-Id: I2659fe985f1a815352db67faa3788fb3c239ea66

Change-Id: I8a83f8ec4db91cd9d5074004ce6fdf7c55b47f5d

Linux doesn't support X2APIC without IRQ remapping unless it recognises
that it's running as a guest. For now we can pretend to be KVM to
trigger hypervisor detection.

ZX-1257 #done

Change-Id: I66094b65f4299c5a6a51eab8ce24f419e05d23f7

Change-Id: I8e490d0a17af3e7e1d23527bc79421288b74f50b

- The zxrio_describe_t object replaces the zxrio_object_t
with a FIDL2 compatible format.
- Flip the usage of ZX_FS_FLAG_PIPELINE to be implied by
default. "ZX_FS_FLAG_DESCRIBE" may be used to request a
description object.
- ZXRIO_DESCRIBE can be trivially implemented using the
description object -- it currently isn't, because there
would be no clients through RIO.

ZX-1360 #comment In Progress

Change-Id: I1c7cad185af468b26c770dfd092382814a436c04

If a server is unable to handle a request at the moment, but may
be able to do so if the client waits, it can respond with a
TFTP_ERR_CODE_BUSY error packet. This behavior is induced by a
response of TFTP_ERR_SHOULD_WAIT from a file_open_read() or
file_open_write() call.

ZX-1521

Change-Id: I2551407052d4a33a7e6b6eebcf45d8f4b28b9dab

When using tftp, recognize that a TFTP_ERR_SHOULD_WAIT result indicates
that we should retry transmission after a delay.

ZX-1521

Change-Id: I7a8bb039b8f6b39236c93d6c9ec9a9a4290e06dc

As the tool now adds blobs concurrently, on high CPU Mac systems in
particular, RLIMIT_NOFILE starts far too low, and we begin to fail against
the soft limit.

Change-Id: I06b1d42d0cf4e41182ca0b1652ccbad00fc7e6cf

This is designed to be included multiple times for different X macros,
so name it accordingly.

Change-Id: I1899577dfffb828d3845bc0d6ebc4ebd0f79a179

Key idea is to replace SetBss(...) with
ConfigureBss(wlan_bss_config_t), which can be used for both, AP and
client mode.

Change-Id: I3283fc57296ee5e4c80c44f158bb6435daf2d6d9

…eporting

Change-Id: Ie27a04dcd83d639ca2d4a029773c4d6bdadee1e3

These are type-safe versions of zx_time_t and zx_duration_t.

Change-Id: Iabf99cb272e488a91204aa307fc015d0b65ce017

Change-Id: I801617022b98c66ecd8d3999cb72fe3783a82284

Change-Id: If45414abe19519a20b08b2916b052ea8cc2f9c6b

This was overlooked because this build doesn't occur in zircon itself,
but only when it rolls up into Garnet.

Change-Id: I244ad653352992d8fd7f94b121a0cd0141036157

This change hoists out the code for generating the part of the
output path that is independent of the test binary's name, since
it's unnecessary to do this every time. It also resolves the
path to the test binary using realpath to completely eliminate
any chance of output name collisions via relative paths (e.g.
runtests -o /data /system/test ./system/test).

Change-Id: Ibe014745944df3e836b772564dff6c36c2469728

ZX-1545 #comment In Progress

Change-Id: I456c71c1b093791d1e21855849d566726a4c0544

The commit adding error_reporter.{cpp,h} neglected to add them to the
BUILD.gn file.

Change-Id: Ifd71505d486c3e7d798781baf393ba8a75e72500

…oo large a result

ZX- 1540 #done

Change-Id: I6fcb49d78fdc0251416fd56a9b5bcefcf40d5bf1

Add bit-twiddling to turn promiscuous on and off for AX88179, AX887722b,
Realtek 8111, and intel-ethernet drivers.

Test: builds; ethernet still works.

NET-9

Change-Id: I3683685c8e2969096621599f92e6d01f68fed404

zx-1553 #done

Change-Id: Ief36d7195a347cbdaa1995273dbaa941438438e0

When paving, store incoming tftp data to a buffer so that the ethernet
receive buffers don't fill up. This introduces an additional thread of
execution - one to receive the eth data and one to write it out to the
paver process.

ZX-1521

Change-Id: Ic2a05c195a872deed8975727ee85526072d932f9

Save and restore FPU state if the guest attempted to use the FPU.
Additionally save PAR_EL1 when doing a world-switch.

Change-Id: I162dbab429a669008a43a6a7ac687615fb7d918f

We don't need the ISB before altering FPU state.

Change-Id: I678845953b470fc1b4c66cad850e4f3a12e00e88

Change-Id: I4f8eb7d7d9ae0f205d1cb623856d11b8a8810b61

Change-Id: Ib4fa78a634971cf49f33938153780fa01d159a03

This removes the percpu.stats related to exceptions
and adds fine granded kernel counters accessible via
"k counters".

The arm counters are a subset of the x64 counters.

Change-Id: I7d87f9309d43ea58da3cc60fd3a5da8a987672b1

It is now possible to wait for multiple interrupts on a single interrupt handle.
The zx_interrupt_bind() syscall is used to bind an interrupt vector to
an interrupt handle, allowing up to 62 interrupt vectors per handle.

The zx_interrupt_get_timestamp() syscall can now be used to retrieve a
timestamp for an interrupt.

Virtual interrupts are now supported. zx_interrupt_bind() can be used to
designate an interrupt slot as a virtual interrupt, and the virtual interrupt
can be signaled using the zx_interrupt_signal() syscall.

The zx_interrupt_complete() syscall has been removed.
Instead, zx_interrupt_wait() now handles unmasking the interrupt
at the right time depending on whether the interrupt is edge or level triggered.

Slot ZX_INTERRUPT_SLOT_USER is reserved as a virtual interrupt, and is typically
used with zx_interrupt_signal() to unblock interrupt threads so they can exit
at driver shutdown or to perform periodic work.

Currently PCI interrupt handles can only be bound to a single interrupt vector,
but it is possible to bind virtual interrupts to PCI interrupt handles.

ZX-1469 #done
ZX-1471 #done

Change-Id: I48c14183c1b9ecd437c8b1ce0d25326b7532e973

Especially during tests we add/remove a lot of block devices.
Let's be less verbose about that.

Change-Id: I3a0f8f77afad6896b9d45db7c1df5f42d00595d3

Change-Id: If009e45c7f3b90fca19f0bb2186cf13c39296a9c

ZX-1129 #comment

Change-Id: I57c6a7cd26e30b7e0252a15b3684ab9d03b9b083

These were converted from google docs automatically, then fixed up
manually.

Note these may not be up to date. kulakowski@ will work on updating them
when he gets back from vacation.

TO-321

Change-Id: If974bcb00a46bfe605a1005287b7b31df717f427

There's only one chipset, which needs to be in promisc mode if
anyone wants it. Track who's asked for it so we know when to
turn it on and off.

Test: builds; ethernet still works; netdump -p receives more on Ax88179

NET-9

Change-Id: Idba2ae3d7a49768d3f664528dbd51de1c1f42117

Use a refcount for freeing paver buffer so that if the paver copy
thread ends before transmission is complete netsvc does not write
to free'd memory.

Restructure file_close/end_connection calls so that files are not
doubly-closed.

Don't mexec when the paver is still operating.

ZX-1521

Change-Id: I2f32a5d18cb904aadfe5bb63646b0a935ff1d54e

This patch starts to introduce the C++ FIDL types. These types are views
into the underlying C types that are used by the encode and decode
functions. For decoding, these views are ultimately on the underlying
message buffer, similar to the C types. For encoding, the underlying C
types are stored in a builder and manipulated through the C++ views.

Change-Id: I21b8e2be687c2cfed57d395898e31e3d16bc9961

If the block core device we bind to supports the new protocol:
- support it ourselves and pass query() and queue() ops through 1:1
- our iotxn_queue() will convert to block_queue() for read/write support
- BlockServer::Queue() will use block_queue()

Due to concerns about the 0-based length field and its possibly
constrained sized, upgrade length to 1-based, uint32 block count.

Make use of an anonymous union to make command-specific argument
access a bit less clunky.

Reject block-core devices with block sizes of <512 or non-power-of-two

Fix zxcrypt test to stop trying to create impossible (block size 3)
ramdisks, which now upset the block middle layer.

Update ramdisk test to reflect stricter block alignment requirements.

Change-Id: Iec0c95341d5935494253facd5074ce4227c9b83c

Change-Id: I30a9ec7c43a53ba4ea3b28026e1654231efde823

If we bind to a device using the new protocol:
- support the new protocol ourselves
- pass block_queue() through to parent after validating
  parameters and adjusting starting block

Change-Id: I7befd5c5c65f3580d15611932dc3d9f2dd463adc

If we bind to a device using the new protocol:
- support the new protocol ourselves
- pass block_queue() through to parent after validating
  parameters and adjusting starting block

Change-Id: Ie2565fb3030e789ac5637b3897bceb9bc868ff9d

Change-Id: I3e53dcd4449f83790675262c1249702f24ca70f4

Change-Id: I537be0e908d81a2dd621770eb0c39988667b7317

Change a few things to make it slightly clearer:
* Rename el2_mexec to el2_hvc_mexec
* Move the test for an FPU trap into el2_guest_exit (reducing
  instructions in the vector table)
* Reorder the test for FPU state restore to be easier to read

Change-Id: Id9d1861356f545a30e869ca1e4c67e0e80b5a179

When we exit for a WFI, we should wait until the timer expires and the
re-enter on the next instruction. But we should not inject a timer
interrupt.

This fixes issues with Linux guests on arm64 once they enter
arch_cpu_idle.

Change-Id: I67db6ce7fcc03800f55100f209a8b0ae85e4777a

This will make it much easier to test virtio drivers by implementing a
testing backend that overrides these functions to avoid syscalls.

Change-Id: I926fbd4434b50021d10d35d0bb831d2833efabaf

Change-Id: I38c86d78a46b91e0156ee7a3a7c09253f09eb501
NB: This will break Garnet, fix and manual roll forward to follow.

We should reset the virtual timer when we exit the guest, otherwise it
will fire when we are in the host.

Change-Id: Ia00d66f318b67d1c079b90d765a7bb7c0ac6495f

Zircon hypervisor disguises itself as KVM, which makes Linux kernel
believe that kvmclock is available and try to use it (see
kvmclock_init function in /arch/x86/kernel/kvmclock.c in the Linux
kernel code).

This CL changes CPUID handling so that it reports all KVM features
as unsupported, at least for now.

Change-Id: I081a82df9242e47d63ebd11bbb400f17c588a7e4

This option prevents the fshost from mounting any filesystems
(same as its existing behaviour when netsvc.netboot=true), which
can be useful for running filesystem tests that create filesystems
which the fshost might attempt to "steal."

Change-Id: I5fe6582fa97186780bdd1d08b5b116eb1fab7c4f

Capture START_UP IPIs and forward the instruction pointer and APIC ID to
userland so it can bring up another VCPU.

ZX-1246 #comment

Change-Id: I624d87eb6031492dc7125750e28de7c4a1c1662c

An instance cannot be reopened, so prefix the path with '@' to make it
clear that opening the returned path will not give the same instance.

ZX-1019 #done

Change-Id: Icc6b3cb414feb5192a3878b9d6c25a2df35f32c8

Change-Id: I51bb70dccb5eb812c6151883597a0c720946fb51

Clang is unhappy with handle_vcpu being unused when we build Garnet.
It's safe to not hide its use behind an #ifdef, as arm64 will not
currently send down the relevant packet.

Change-Id: I936fa150799f4962a41a4c197d16d8031ffad535

Allow both the virtual and the physical timer to be specified, with the
physical timer only being preferred if we booted at EL2 or above.

Change-Id: I32f867eb56cdc4556d3dddd69a0fef7d4302a2b6

This is so we can specify the ramdisk to use when we're building more
than just Zircon.

Change-Id: I1bec3be35378eaedca0664d1d7c4204e470cef48

Change-Id: I5163b9a40f13d3d61cccfe9423aa3f76176b164a

Add the ability to run a function and assert that it did crash in
addition to the ability to assert that it did.  This lets me write
tests where I have an operation which might crash (because the
code-under-test is broken), but should not, and still terminate the
test cleanly with a clean ASSERT log message after catching the
exception.  Afterwards, other tests in the test case can continue to
run even if this specific test failed.

Change-Id: I7778c07cba16233252d20de1dd9c9bca9fa960a0

When ring->current == ring->dequeue_ptr, we would always return 255
(empty ring), but if we enqueue really large requests, ring->current
may loop back around and we should return 0.

Change-Id: Ifeb903f032569f296cdff98881800c8e05598a28

This change fixes the emitted JSON to include commas between tests
from different directories, which was previosuly not handled.

Change-Id: I78ea84db61de98ea85dcafa3bb8d9ac4cb62de4a

Each video frame will be stored in the data ring buffer shared
with the client.

There will also be a metadata ring buffer, which will have an
entry for each frame.

Clients will be notified when n frames are available. They can
then lock those frames and copy the metadata / data from
the ring buffers.

ZX-1129 #comment

Change-Id: I3fd3de1cee021e188c993a1d7be4accf0d14037c

- do the "in list" check inside the spinlock
- switch back to using a single spinlock so "in list" check is actually safe
- allow to reuse the dpc_t* object in a non-racy way
- return ZX_ERR_ALREADY_EXISTS from dpc_queue if the dpc is already queued.
  this is necessary for using dpc with reference counting.

Change-Id: I2942836d90e462bd78861eb47fe6690a967e08f8

Change-Id: I811979edd080e500c499c6d25657b6d15e2d1712

We want to use bitmap in Garnet and need a target.

Change-Id: I66844f8b2468a8ef58dcf9b7829a3a1ce255bd89

Now it only prints the right conters in
the prefix match or none. In the case of an
exact match just that counter is printed

Also the output formating is more consistent.

Change-Id: If29a5fe60b763c057ac62056e08de2dd512acb2a

dev/power.h #includes itself, which is unnecessary.

I found this when looking for #include cycles.

Change-Id: Ic38499cfd7fbb18776173cd5ea476fc4641d39e4

The older link no longer resolves.

Change-Id: Ida9cc1eacf3b44a81aea7edffde59ea76eb79e00

…in the midst of cancelling all the transfers. The comment in xhci_cancel_transfers talks about the same.

Change-Id: I6c0bed66c33c65fb2f73d0116b0ea2e3c8c14c63

Change-Id: Ib0f2d408676787299a744d68133a010270df500a

Change-Id: I5efaa8f628ea28526df0ee5f29589db83d3e57d3

This is to allow build system work to proceed in parallel with source
generation. It will be unwound when the real thing lands.

Change-Id: Ia2591c87f89101ea6a3d162fb9240b9f7e76f85d

This is so we can modify them in PhysMemFake for testing.

Change-Id: I610a3154209a2de6459c2a48d99b4b49bca899af

Test: /system/test/sys/ethernet-test passes

NET-9 #done

Change-Id: I008c40cf457ef817b071e5c2dff72a9d9f817d57

Change-Id: I76b00f2de3aa0c3422e56fcfc03a6964ac58e7ab

All transmissions are suspended unless reserved bit 28 of TCTL
is set. For minimum disruption, leave all TCTL reserved bits
unchanged when writing to them.

ZX-1516

Change-Id: I6c43b78ec5f276822d1223cd2122ed40785636d3

We need to be able to build this library from Garnet to write the C++
bindings.

Change-Id: I24359f9a3f2727e15923c0303bea7bc6ded5a596