- Don't change RNDIS RSS configuration for RSS key sysctl, if the
interface is not capable of RSS yet.
- Don't change RSS indirect table (both cached one and RNDIS RSS
configuration), if the interface is not capable of RSS yet.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7924
And don't allow capability changes during reinitialization, which
breaks too much static configuration.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7922
NVS and NDIS version change would break too much assumption and static
configuration.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7919
drivers.
The BMIPS32/BMIPS3300 cores use a register layout distinct from the MIPS74K
core, and are only found on siba(4) devices.
Reviewed by: mizhka
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7791
While I'm here, add comment along the attach DEVMETHOD.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7874
* hard code a noise floor of -96 for now. The noise floor update code returns
some "interesting" values that I can't map to anything useful right now.
* Ensure a default noise floor is set - otherwise the initial scan results
have a noise floor of '0'.
* Fix up the RSSI calculation to be correctly relative to the noise floor.
The RSSI routines return an absolute value in dBm - so fix this up.
* Cap RSSI values appropriately.
* Ensure we pass in a 1/2 dB unit value in to net80211.
Tested:
* Intel 7260, STA mode
iwm0: <Intel Dual Band Wireless AC 7260> mem 0xf1400000-0xf1401fff irq 17 at device 0.0 on pci2
iwm0: hw rev 0x140, fw ver 16.242414.0, address xx:xx:xx:xx:xx:xx
In r304870, refcount handling was lifted out into a common OTP/SPROM code
path, but the refcount assertions in chipc_disable_sprom_pins() were not
updated accordingly; this triggered an assertion on BCM4331 devices when
releasing a SPROM pin reservation.
Approved by: adrian (mentor, implicit)
Make virtio_console(4) create `/dev/vtcon/<port_name>` alias pointing
to /dev/ttyVx.y upon receiving PORT_NAME (id = 7) event over the control
queue.
Approved by: trasz
MFC after: 1 month
Sponsored by: iXsystems, Inc.
Differential Revision: https://reviews.freebsd.org/D7182
When mlx5e_sq_xmit() returns an error code and the mbuf pointer is set,
we should not free the mbuf, because the caller will keep the mbuf in
the drbr. Make sure the mbuf pointer is correctly set upon function
exit.
Sponsored by: Mellanox Technologies
MFC after: 1 week
The hardware MTU size can't be set to a value less than 1500 bytes due
to side-band management support. Allow setting the software MTU size
below 1500 bytes, thus creating a mismatch between hardware and
software MTU sizes.
Sponsored by: Mellanox Technologies
MFC after: 1 week
Try to reuse code to setup sendqueues when possible by making some static
functions global. Further split the mlx5e_close_sq_wait() function to
separate out reusable parts.
Sponsored by: Mellanox Technologies
MFC after: 1 week
This change also reduces the size of the mlx5e_sq structure so that the last
queue_state element will fit into the previous cacheline and then the mlx5e_sq
structure becomes one cacheline less for amd64.
Sponsored by: Mellanox Technologies
MFC after: 1 week
Make some functions and structures global to allow for code reuse
when creating rate limiting sendqueues.
Sponsored by: Mellanox Technologies
MFC after: 1 week
Move setting of CQ moderation mode together with the other
CQ moderation parameters. Pass completion event vector as
a separate argument to mlx5e_open_cq(), because its value is
different for each call. Pass mlx5e_priv pointer instead of
mlx5e_channel pointer so that code can be used by rate
limiting sendqueues.
Sponsored by: Mellanox Technologies
MFC after: 1 week
This change allows for reusing the transmit path for so called
rate limited senqueues. While at it optimise some pointer lookups
in the fast path.
Sponsored by: Mellanox Technologies
MFC after: 1 week
- Add new firmware commands and update existing ones.
- Add more firmware related structures and update existing ones.
- Some minor fixes, like adding missing \n to some prints.
Sponsored by: Mellanox Technologies
MFC after: 1 week
come up as 't6nex' nexus devices with 'cc' ports hanging off them.
The T6 firmware and configuration files will be added as soon as they
are released. For now the driver will try to work with whatever
firmware and configuration is on the card's flash.
Sponsored by: Chelsio Communications
mutexes or using any callouts when active.
Trying to lock a mutex when KDB is active or the scheduler is stopped
can result in infinite wait loops. The same goes for calling callout
related functions which in turn lock mutexes.
If the USB controller at which a USB keyboard is connected is idle
when KDB is entered, polling the USB keyboard via USB will always
succeed. Else polling may fail depending on which state the USB
subsystem and USB interrupt handler is in. This is unavoidable unless
KDB can wait for USB interrupt threads to complete before stalling the
CPU(s).
Tested by: Bruce Evans <bde@freebsd.org>
MFC after: 4 weeks
- ifnet.if_mtu does not require explicit setting.
- ifnet.if_hdrlen must be set after ether_ifattach().
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7873
It is available on both stable/10 and stable/11. This eases future MFCs
to stable/10.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7872
This paves the way for the dynamic RSS key and indirect table setting.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7864
This paves the way for further attach/detach code reorganization.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7858
* Don't do RTS/CTS - experiments show that we get ACK frames for each of them
and this ends up causing the timestamps to look all funny.
* Set the HAL_TXDESC_POS bit, so the AR9300 HAL sets up the hardware to return
location and CSI information.
evdev is a generic input event interface compatible with Linux
evdev API at ioctl level. It allows using unmodified (apart from
header name) input evdev drivers in Xorg, Wayland, Qt.
This commit has only generic kernel API. evdev support for individual
hardware drivers like ukbd, ums, atkbd, etc. will be committed later.
Project was started by Jakub Klama as part of GSoC 2014. Jakub's
evdev implementation was later used as a base, updated and finished
by Vladimir Kondratiev.
Submitted by: Vladimir Kondratiev <wulf@cicgroup.ru>
Reviewed by: adrian, hans
Differential Revision: https://reviews.freebsd.org/D6998
Trying to build a MIPS platform that uses INTRNG needs this
for this to work right in gpiobusvar.h :
#ifdef INTRNG
struct intr_map_data_gpio {
struct intr_map_data hdr;
...
};
#endif
* change the HT_RC_2_MCS to do MCS0..23
* Use it when looking up the ht20/ht40 array for bits-per-symbol
* add a clk_to_psec (picoseconds) routine, so we can get sub-microsecond
accuracy for the math
* .. and make that + clk_to_usec public, so higher layer code that is
returning clocks (eg the ANI diag routines, some upcoming locationing
experiments) can be converted to microseconds.
Whilst here, add a comment in ar5416 so i or someone else can revisit the
latency values.
The GUID string provided by hypervisor has leading and trailing braces,
while our GUID string does not have braces at all. Both braces should
be ignored, when the GUID strings are compared.
Submitted by: Hongjiang Zhang <honzhan microsoft com>
Modified by: sephe
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7809
While I'm here, pull up the channel callback related code too.
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7805
The cxgbev/cxlv driver supports Virtual Function devices for Chelsio
T4 and T4 adapters. The VF devices share most of their code with the
existing PF4 driver (cxgbe/cxl) and as such the VF device driver
currently depends on the PF4 driver.
Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf
PCI device driver that attaches to the VF device. It then creates
child cxgbev/cxlv devices representing ports assigned to the VF.
By default, the PF driver assigns a single port to each VF.
t4vf_hw.c contains VF-specific routines from the shared code used to
fetch VF-specific parameters from the firmware.
t4_vf.c contains the VF-specific PCI device driver and includes its
own attach routine.
VF devices are required to use a different firmware request when
transmitting packets (which in turn requires a different CPL message
to encapsulate messages). This alternate firmware request does not
permit chaining multiple packets in a single message, so each packet
results in a firmware request. In addition, the different CPL message
requires more detailed information when enabling hardware checksums,
so parse_pkt() on VF devices must examine L2 and L3 headers for all
packets (not just TSO packets) for VF devices. Finally, L2 checksums
on non-UDP/non-TCP packets do not work reliably (the firmware trashes
the IPv4 fragment field), so IPv4 checksums for such packets are
calculated in software.
Most of the other changes in the non-VF-specific code are to expose
various variables and functions private to the PF driver so that they
can be used by the VF driver.
Note that a limited subset of cxgbetool functions are supported on VF
devices including register dumps, scheduler classes, and clearing of
statistics. In addition, TOE is not supported on VF devices, only for
the PF interfaces.
Reviewed by: np
MFC after: 2 months
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D7599
If a packet contains the Ethernet header (14 bytes) in the first mbuf
and the payload (IP + UDP + data) in the second mbuf, then the attempt
to fetch the l3hdr will return a NULL pointer. The first loop iteration
will drop len to zero and exit the loop without setting 'p'. However,
the desired data is at the start of the second mbuf, so the correct
behavior is to loop around and let the conditional set 'p' to m_data of
the next mbuf (and leave offset as 0).
Reviewed by: np
Sponsored by: Chelsio Communications
Due to reading initialized variable, FIS receive area was always allocated
as 256 bytes, suitable for command-based switching, instead of 4096 bytes,
required for FIS-based switching. This caused memory corruption in case of
port multipliers used on FBS-capable HBAs (Marvell).
MFC after: 1 week
Add routines to trigger a function level reset (FLR) of a PCI-express
device via the PCI-express device control register. This also includes
support routines to wait for pending transactions to complete as well
as calculating the maximum completion timeout permitted by a device.
Change the ppt(4) driver to reset pass through devices before attaching
to a VM during startup and before detaching from a VM during shutdown.
Reviewed by: imp, wblock (earlier version)
MFC after: 1 month
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D7751
This driver supports two bindings:
- cpufreq-dt: systems which share clock and voltage across all CPUs
- arm_big_little_dt: systems which share clock and voltage across all
CPUs in a single cluster
Reviewed by: andrew, imp
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D7741
When the I/O MMU is active in bhyve, all PCI devices need valid entries
in the DMAR context tables. The I/O MMU code does a single enumeration
of the available PCI devices during initialization to add all existing
devices to a domain representing the host. The ppt(4) driver then moves
pass through devices in and out of domains for virtual machines as needed.
However, when new PCI devices were added at runtime either via SR-IOV or
HotPlug, the I/O MMU tables were not updated.
This change adds a new set of EVENTHANDLERS that are invoked when PCI
devices are added and deleted. The I/O MMU driver in bhyve installs
handlers for these events which it uses to add and remove devices to
the "host" domain.
Reviewed by: imp
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D7667
Other files including pci_host_generic.h failed to compile
due to missing declaration of enum pci_id_type.
Obtained from: Semihalf
Submitted by: Michal Stanek <mst@semihalf.com>
Sponsored by: Annapurna Labs
Reviewed by: wma
Differential Revision: https://reviews.freebsd.org/D7561
Split usbd_xfer_status() check:
- Check xfer length: must be longer, than Rx descriptor size.
- Check frame size: must be shorter than xfer length.
- Discard too short frames.
Tested with WUSB54GC, STA/MONITOR modes.
This adds bhnd(4) bus-level support for querying backplane interrupt vector
routing, and delegating machine/bridge-specific interrupt handling to the
concrete bhnd(4) driver implementation.
On bhndb(4) bridged PCI devices, we provide the PCI/MSI interrupt directly
to attached cores.
On MIPS devices, we report a backplane interrupt count of 0, effectively
disabling the bus-level interrupt assignment. This allows mips/broadcom
to temporarily continue using hard-coded MIPS IRQs until bhnd_mips PIC
support is implemented.
Reviewed by: mizhka
Approved by: adrian (mentor, implicit)
Broadcom Intensi-fi chipsets provided a common set of IP cores; on PCI/PCIe
devices, the USB11 host controller is left floating.
Approved by: adrian (mentor, implicit)
This patch adds driver implementation for BHND USB core. Driver has been
imported from ZRouter project with small adaptions for FreeBSD 11.
Also it's enabled for BroadCom MIPS74k boards by default. It's fully tested
on Asus boards (RT-N16: external USB, RT-N53: USB bus between SoC and WiFi
chips).
Reviewed by: adrian (mentor), ray
Approved by: adrian (mentor)
Obtained from: ZRouter
Differential Revision: https://reviews.freebsd.org/D7781
by reviving the SX control request lock and refining which lock
protects the common scratch area in "struct usb_device".
The SX control request lock was removed by r246759 because it caused a
lock order reversal with the USB enumeration lock inside
usbd_transfer_setup() as a function of r246616. It was thought that
reducing the number of locks would resolve the LOR, but because some
USB device drivers use usbd_do_request_flags() inside callback
functions, like in taskqueues, a deadlock may occur when these are
drained from device_detach(). By restoring the SX control request
lock usbd_do_request_flags() is allowed to complete its execution
when a USB device driver is detaching. By using the SX control request
lock to protect the scratch area, the LOR introduced by r246616 is
also resolved.
Bump the FreeBSD version while at it to force recompilation of all USB
kernel modules.
Found by: avos@
MFC after: 1 week
sys/dev/usb/serial/uplcom.c:543:29: error: implicit conversion from 'int' to 'int8_t' (aka 'signed char') changes value from 192 to -64 [-Werror,-Wconstant-conversion]
if (uplcom_pl2303_do(udev, UT_READ_VENDOR_DEVICE, UPLCOM_SET_REQUEST, 0x8484, 0, 1)
~~~~~~~~~~~~~~~~ ^~~~~~~~~~~~~~~~~~~~~
sys/dev/usb/usb.h:179:53: note: expanded from macro 'UT_READ_VENDOR_DEVICE'
#define UT_READ_VENDOR_DEVICE (UT_READ | UT_VENDOR | UT_DEVICE)
~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
This is because UT_READ is 0x80, so the int8_t argument is wrapped to a
negative value. Fix this by using uint8_t instead.
Reviewed by: imp, hselasky
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D7776
This fixes bhnd(4) nvram handling on devices that map SPROM CSRs via PCI
configuration space.
The probe method previously required that a bhnd(4) device be attached to the
parent bridge; now that the bhnd_nvram device is always attached first, this
unnecessary sanity check always failed.
Approved by: adrian (mentor, implicit)
On BCM4321 chipsets, both PCI and PCIe cores are included, with one of
the cores potentially left floating.
Since the PCI core appears first in the device table, and the PCI
profiles appear first in the resource configuration tables, this resulted in
incorrectly matching and using the PCI/v1 resource configuration on PCIe
devices, rather than the correct PCIe/v1 profile.
Approved by: adrian (mentor, implicit)
Adds support for probing and initializing bhndb(4) bridge state using
the bhnd_erom API, ensuring that full bridge configuration is available
*prior* to actually attaching and enumerating the bhnd(4) child device,
allowing us to safely allocate bus-level agent/device resources during
bhnd(4) bus enumeration.
- Add a bhnd_erom_probe() method usable by bhndb(4). This is an analogue
to the existing bhnd_erom_probe_static() method, and allows the bhndb
bridge to discover the best available erom parser class prior to newbus
probing of its children.
- Add support for supplying identification hints when probing erom
devices. This is required on early EXTIF-only chipsets, where chip
identification registers are not available.
- Migrate bhndb over to the new bhnd_erom API, using bhnd_core_info
records rather than bridged bhnd(4) device_t references to determine
the bridged chipsets' capability/bridge configuration.
- The bhndb parent (e.g. if_bwn) is now required to supply a hardware
priority table to the bridge. The default table is currently sufficient
for our supported devices.
- Drop the two-pass attach approach we used for compatibility with bhndb(4) in
the bhnd(4) bus drivers, and instead perform bus enumeration immediately,
and allocate bridged per-child bus-level resources during that enumeration.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7768
This defines a new bhnd_erom_if API, providing a common interface to device
enumeration on siba(4) and bcma(4) devices, for use both in the bhndb bridge
and SoC early boot contexts, and migrates mips/broadcom over to the new API.
This also replaces the previous adhoc device enumeration support implemented
for mips/broadcom.
Migration of bhndb to the new API will be implemented in a follow-up commit.
- Defined new bhnd_erom_if interface for bhnd(4) device enumeration, along
with bcma(4) and siba(4)-specific implementations.
- Fixed a minor bug in bhndb that logged an error when we attempted to map the
full siba(4) bus space (18000000-17FFFFFF) in the siba EROM parser.
- Reverted use of the resource's start address as the ChipCommon enum_addr in
bhnd_read_chipid(). When called from bhndb, this address is found within the
host address space, resulting in an invalid bridged enum_addr.
- Added support for falling back on standard bus_activate_resource() in
bhnd_bus_generic_activate_resource(), enabling allocation of the bhnd_erom's
bhnd_resource directly from a nexus-attached bhnd(4) device.
- Removed BHND_BUS_GET_CORE_TABLE(); it has been replaced by the erom API.
- Added support for statically initializing bhnd_erom instances, for use prior
to malloc availability. The statically allocated buffer size is verified both
at runtime, and via a compile-time assertion (see BHND_EROM_STATIC_BYTES).
- bhnd_erom classes are registered within a module via a linker set, allowing
mips/broadcom to probe available EROM parser instances without creating a
strong reference to bcma/siba-specific symbols.
- Migrated mips/broadcom to bhnd_erom_if, replacing the previous MIPS-specific
device enumeration implementation.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7748
sys/dev/cxgb/cxgb_sge.c:2873:44: error: implicit conversion from 'int'
to 'char' changes value from 128 to -128 [-Werror,-Wconstant-conversion]
*mtod(m, char *) = CPL_ASYNC_NOTIF;
~ ^~~~~~~~~~~~~~~
This is because CPL_ASYNC_NOTIF is 0x80, so the plain char argument is
wrapped to a negative value. Fix this by using uint8_t instead.
Reviewed by: np
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D7772
sys/dev/ppbus/ppb_1284.c:296:46: error: implicit conversion from 'int'
to 'char' changes value from 144 to -112 [-Werror,-Wconstant-conversion]
if ((error = do_peripheral_wait(bus, SELECT | nBUSY, 0))) {
~~~~~~~~~~~~~~~~~~ ~~~~~~~^~~~~~~
sys/dev/ppbus/ppb_1284.c:785:48: error: implicit conversion from 'int'
to 'char' changes value from 240 to -16 [-Werror,-Wconstant-conversion]
if (do_1284_wait(bus, nACK | SELECT | PERROR | nBUSY,
~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~
sys/dev/ppbus/ppb_1284.c:786:29: error: implicit conversion from 'int'
to 'char' changes value from 240 to -16 [-Werror,-Wconstant-conversion]
nACK | SELECT | PERROR | nBUSY)) {
~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~
This is because nBUSY is 0x80, so the plain char argument is wrapped to
a negative value. Fix this in a minimal fashion, by using uint8_t in a
few places.
Reviewed by: emaste
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D7771
transfers.
The Initiator and Target both perform zero copy receive for transfers
greater than or equal to this threshold.
Sponsored by: Chelsio Communications
routines available in t4_tom to manage the iSCSI DDP page pod region.
This adds the ability to use multiple DDP page sizes to the iSCSI
driver, among other improvements.
Sponsored by: Chelsio Communications
important detail that sc_cngetc() now opens and closes the keyboard
on every call again. This was moved from sc_cngetc() to scn_cngrab/
ungrab() in r228644, but the change wasn't quite complete. After
fixes for nesting in kbdd_poll() in ukbd and kbdmux, these opens
and closes should have no significant effect if done while grabbed.
They fix unusual cases when cngetc() is called while not grabbed.
This commit is the main fix for screen locking in sc_cnputc():
detect deadlock or likely-deadlock and handle it by buffering the
output atomically and printing it later if the deadlock condition
clears (and sc_cnputc() is called).
The most common deadlock is when the screen lock is held by ourself.
Then it would be safe to acquire the lock recursively if the console
driver is calling printf() in a safe context, but we don't know when
that is. It is not safe to ignore the lock even in kdb or panic mode.
But ignore it in panic mode. The only other known case of deadlock
is when another thread holds the lock but is running on a stopped CPU.
Detect that case approximately by using trylock and retrying for 1000
usec. On a 4 GHz CPU, 100 usec is almost long enough -- screen switches
take slightly longer than that. Not retrying at all is good enough
except for stress tests, and planned future versions will extend the
timeout so that the stress tests work better.
To see the behaviour when deadlock is detected, single step through
sctty_outwakeup() (or sc_puts() to start with deadlock). Another
(serial) console is needed to the buffered-only output, but the
keyboard works in this context to continue or step out of the
deadlocked region. The buffer is not large enough to hold all the
output for this.
The purpose of BHND_PMU_{GET,SET}_BITS macro is to transform values from/into
register format. SET macro shifts value to left and applies filter mask.
GET macro applies filter mask and then shifts value to right.
Reviewed by: landonf, adrian (mentor)
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D7721
The variables that are extern in the netmap header file should be
defined in ixl_txrx.c (the file that is included in both ixl(4)/ixlv(4),
not in the main driver source files.
Reported by: ed@, dim@, ngie@
Keyboard input needs Giant locking, and that is not possible to do
correctly here. Use mtx_trylock() and proceed unlocked as before if
we can't acquire Giant (non-recursively), except in kdb mode don't
even try to acquire Giant. Everything here is a hack, but it often
works. Even if mtx_trylock() succeeds, this might be a LOR.
Keyboard input also needs screen locking, to handle screen updates
and switches. Add this, using the same simplistic screen locking
as for sc_cnputc().
Giant must be acquired before the screen lock, and the screen lock
must be dropped when calling the keyboard driver (else it would get a
harmless LOR if it tries to acquire Giant). It was intended that sc
cn open/close hide the locking calls, and they do for i/o functions
functions except for this complication.
Non-console keyboard input is still only Giant-locked, with screen
locking in some called functions. This is correct for the keyboard
parts only.
When Giant cannot be acquired properly, atkbd and kbdmux tend to race
and work (they assume that the caller acquired Giant properly and don't
try to acquire it again or check that it has been acquired, and the
races rarely matter), while ukbd tends to deadlock or panic (since it
does the opposite, and has other usb threads to deadlock with).
The keyboard (Giant) locking here does very little, but the screen
locking completes screen locking for console mode except for not
detecting or handling deadlock.
- By default, adjust time upon SYNC request. It can be disabled
through hw.hvtimesync.ignore_sync_req. SYNC request will be
sent by hypervisor the host is resumed, rebooted, etc.
- By default, adjust time upon SAMPLE request, if there is 100ms
difference between VM time and hypervisor time. This can be
disabled through hw.hvtimesync.sample_drift.
And nuke the unnecessary task, since channel callback is running
in a Hyper-V taskqueue nowadays.
Submitted by: YanZhe Chen <t-yachen microsoft com>
Discussed with: Dexuan Cui <decui microsoft com>, Hongjiang Zhang <honzhan microsoft com>, sephe
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D7707
