Hardware buffer management entries are not used yet by FreeBSD.
They were added for compliance with Linux Armada 38x device tree
representation and will be used in future network support.
Submitted by: Bartosz Szczepanek <bsz@semihalf.com>
Obtained from: Semihalf
Sponsored by: Stormshield
Differential revision: https://reviews.freebsd.org/D8179
- recognize ports and vlangroups based on DTS file
- support multi-chip addresing mode (required in upcoming
Armada-388-Clearfog support)
- refactor attachment function
Each port in 'dsa' node should have 'vlangroup' property. Otherwise,
e6000sw will fail to attach.
Submitted by: Bartosz Szczepanek <bsz@semihalf.com>
Konrad Adamczyk <ka@semihalf.com>
Obtained from: Semihalf
Sponsored by: Stormshield
Differential revision: https://reviews.freebsd.org/D7328
* rename the ieee80211com field for vht mcsinfo to be ic_, not iv;
* add a vht config field, stealing from the spares I left there.
This doesn't change the ABI.
The HT40 channel population logic was "just" doing pairs of channels starting with
the band entry frequency. Trouble is, a lot of the rules start way off at 5120MHz,
which isn't a valid 5GHz channel. Then, eg for HT40U, it would populate:
* (5120,5140)
* (5160,5180)
* (5200,5220)
* (5240,5260)
.. as the HT40U pairs, with the first being the primary channel. Channel 36
is 5180MHz, and since it's not a primary channel here, it wouldn't populate it.
Then, the next HT40U would be 5200/5220, which is highly wrong.
HT40D had the same problem.
So, this just forces that 5GHz HT40 channels start at channel 36 (5180),
no matter what the band edge says. This includes eg doing 4.9GHz channels.
This erm, meant that the HT40 channels for the low band was always wrong.
Oops!
Tested:
* AR9380, STA mode
* AR9344 SoC, AP mode
MFC after: 1 week
Upon each execve, we allocate a KVA range for use in copying data to the
new image. Pages must be faulted into the range, and when the range is
freed, the backing pages are freed and their mappings are destroyed. This
is a lot of needless overhead, and the exec_map management becomes a
bottleneck when many CPUs are executing execve concurrently. Moreover, the
number of available ranges is fixed at 16, which is insufficient on large
systems and potentially excessive on 32-bit systems.
The new allocator reduces overhead by making exec_map allocations
persistent. When a range is freed, pages backing the range are marked clean
and made easy to reclaim. With this change, the exec_map is sized based on
the number of CPUs.
Reviewed by: kib
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D8921
Previously, the stack unwinder tried to locate the start of the function
in each frame by walking backwards until it found an instruction that
modified the stack pointer and then assumed that was the first instruction
in a function. The unwinder would only print a function name if the
starting instruction's address was an exact match for a symbol name.
However, not all functions generated by modern compilers start off functions
with that instruction. For those functions, the unwinder would fail to
find a matching function name. As a result, most frames in a stack
trace would be printed as raw hex PC's instead of a function name.
Stop depending on this incorrect assumption and just use db_printsym()
like other platforms to display the function name and offset for each
frame. This generates a far more useful stack trace.
While here, don't print out curproc's pid at the end of the trace. The
pid was always from curproc even if tracing some other process.
In addition, remove some rotted comments about hardcoded constants that
are no longer hardcoded.
Sponsored by: DARPA / AFRL
There was a single call to stacktrace() under an #ifdef DEBUG to obtain
a stack trace during a fault that resulted in a function pointer to a
printf function being passed to stacktrace_subr() in db_trace.c. The
kernel now has existing interfaces for obtaining a stack trace outside
of DDB (kdb_backtrace(), or the stack_*() API) that should be used instead.
Rather than fix the one call however, remove it since the kernel will
dump a trace anyway once it panics.
Make stacktrace_subr() static, remove the function pointer and change it
to use db_printf() explicitly.
Discussed with: kan
Sponsored by: DARPA / AFRL
Use the trapframe unwinder recently added for kernel stack overflow
panics for frames crossing MipsKernGenException and MipsKernIntr.
This provides more reliably unwinding across nested interrupts and
exceptions in the kernel.
While here, dump the value of the CAUSE and BADVADDR registers when
crossing a trapframe.
Submitted by: rwatson (original version)
Obtained from: CheriBSD
Sponsored by: DARPA / AFRL
The sim_vid, hba_vid, and dev_name fields of struct ccb_pathinq are
fixed-length strings. AFAICT the only place they're read is in
sbin/camcontrol/camcontrol.c, which assumes they'll be null-terminated.
However, the kernel doesn't null-terminate them. A bunch of copy-pasted code
uses strncpy to write them, and doesn't guarantee null-termination. For at
least 4 drivers (mpr, mps, ciss, and hyperv), the hba_vid field actually
overflows. You can see the result by doing "camcontrol negotiate da0 -v".
This change null-terminates those fields everywhere they're set in the
kernel. It also shortens a few strings to ensure they'll fit within the
16-character field.
PR: 215474
Reported by: Coverity
CID: 1009997 1010000 1010001 1010002 1010003 1010004 1010005
CID: 1331519 1010006 1215097 1010007 1288967 1010008 1306000
CID: 1211924 1010009 1010010 1010011 1010012 1010013 1010014
CID: 1147190 1010017 1010016 1010018 1216435 1010020 1010021
CID: 1010022 1009666 1018185 1010023 1010025 1010026 1010027
CID: 1010028 1010029 1010030 1010031 1010033 1018186 1018187
CID: 1010035 1010036 1010042 1010041 1010040 1010039
Reviewed by: imp, sephe, slm
MFC after: 4 weeks
Sponsored by: Spectra Logic Corp
Differential Revision: https://reviews.freebsd.org/D9037
Differential Revision: https://reviews.freebsd.org/D9038
returns memory which must be freed, regardless of the error. Assign
NULL to *buf in case we are not going to allocate any memory due to
invalid mode.
Reported and tested by: pho
Reviewed by: jhb
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks (together with r310638)
Differential revision: https://reviews.freebsd.org/D9042
a symlink and an autofs mount request. The crash was caused by namei()
calling bcopy() with a negative length, caused by numeric underflow:
in lookup(), in the relookup path, the ni_pathlen was decremented too
many times. The bug was introduced in r296715.
Big thanks to Alex Deiter for his help with debugging this.
Reviewed by: kib@
Tested by: Alex Deiter <alex.deiter at gmail.com>
MFC after: 1 month
observed to fix any actual error, but it's the right thing to do
from the correctness point of view.
Tested by: Eugene M. Zheganin <emz at norma.perm.ru>
MFC after: 1 month
Add basic support for A33/R16 that is enough to boot a kernel.
This adds the platform code, padconf data and the new clocks strings.
MFC after: 2 weeks
dangerous. Those wanting data from an mbuf should use DTrace itself to get
the data.
PR: 203409
Reviewed by: hiren
MFC after: 1 week
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D9035
all public firmwares for all chips since the last release (1.15.37.0)
follows (it's a straight copy-paste from the Release Notes for the
12/30/2016 Unified Wire release on Chelsio's website).
T6 Firmware
++++++++++++
Version : 1.16.26.0
Date : 12/28/2016
Fixes
-----
BASE:
- Max number of egress and control queues adjusted to accomodate
co-processor mode queues.
- Fixed intermittent DDR3/4 ECC errors.
- Fixed a traffic stall when ETS BW is configured as 0%.
- Max number of ethctrl queue in VF set to 1.
ETH:
- Added a new config file option 'speed' under port section to set the
port speed. Use only when auto negotiation is off.
- FEC option removed from firmware config file. cxgbtool can be used to
change the fec setting.
- CPL_TX_TNL_LSO cpl handling added in ETH_TX_PKT_VM handler. This fixes
large tunnel tcp packet support for VxLAN.
Version : 1.16.22.0
Date : 12/05/2016
Fixes
-----
BASE:
- fw_port_type updated in fw API to match kernel.org definitions.
- Saved power by disaling unused MAC lanes.
- Configures correct power bin.
- Enhanced DDR4 performance.
- Enabled interrupts.
- Fixed an issue where filter rule for 'unicast hash' is not working.
ETH:
- Disabled auto negotiation by default because most of 100G switches do
not support AN as of today.
- Fixed flow control not getting disabled problem.
- Fixed an issue where port0 doesn't come up sometimes.
- Fixed 10G link not coming up issue.
- Fixed an issue with promiscuous mode when dcbx disabled.
OFLD:
- Fixed a connection stuck issue when abort is received during out of tx
pages backpressure.
ENHANCEMENTS
------------
BASE:
- Added inline TLS mode support.
Version : 1.16.12.0
Date : 11/11/2016
ENHANCEMENTS
------------
BASE:
- Added T6 support.
- Added T6 1G/10G/25G/40G/100G link speeds.
- Added T6 co-processor mode crypto support.
- Added facility to increase link AN+AEC timeout.
OFLD:
- Added support for all T5 offload protocols except FCoE.
iSCSI:
- iscsi completion moderation enabled.
=======================================================================
T5 Firmware
++++++++++++
Version : 1.16.26.0
Date : 12/28/2016
FIXES
-----
BASE:
- Max number of ethctrl queue in VF set to 1.
Version : 1.16.22.0
Date : 12/05/2016
FIXES
-----
BASE:
- Fixed an issue where filter rule for 'unicast hash' is not working.
ETH:
- Fixed an issue with promiscuous mode when dcbx disabled.
ENHANCEMENTS
------------
ETH:
- Added 40G-KR support.
Version : 1.16.12.0
Date : 11/11/2016
FIXES
-----
BASE:
- Fixed multiple issues related with VFs FLR processing.
- Fixed channel assignment based on number of ports in adapter.
- Fixed a crash when VM having PF assigned as passthrough mode is
rebooted.
- Handled 2nd HELLO command from the same PF without seeing BYE from the
same PF and if that is the only PF.
- A warning is printed in firmware log if PCI-E cookie generation is
enabled in serial initialization file.
- Fixed multiple issues related with Filtering.
- Enabled DSGL memory write for iscsi and rdma.
- Added new FW_PARAMS_CMD[DEV] options to retrieve Serial Configuration
and VPD version numbers.
- Fixed an issue where LVDS output was not getting enabled using vpd.
DCBX:
- Fixed DCBX CEE Incorrect class to pririty mapping.
- Fixed incorrect interpretation of DCBX IEEE PFC.
ETH:
- Adjusted the link related delay timings according to the QSFP spec.
- Improved 40G link bringup time with few switches.
OFLD:
- Do not reserve qp/cq if rdma capability is not enabled.
- Fixed an issue where approx 1600+ TOE connections were causing a
firmware fatal error.
FOiSCSI:
- Fixed an issue where unloading foiscsi driver causes mailbox timeout.
ENHANCEMENTS
------------
BASE:
- Added 10G KR/KX support.
- Added T540-BT adapter support.
- Added 4 new rss key modes for PFs and VFs.
OFLD:
- Added new WR FW_RI_FR_NSMR_TPTE_WR to improve fast MR write
performance in RDMA.
Version : 1.16.5.0
Date : 10/26/2016
FIXES
-----
BASE:
- Fixed multiple issues where FLR from multiple VFs can cause firmware
crash.
- Fixed channel assignment based on number of ports in adapter.
- Fixed the HELLO command master force api to handle the 2nd HELLO
correctly without getting BYE from the PF driver.
- Added facility to retrieve Serial configuration and VPD version. Two
new FW_PARAMS_CMD[DEV] options added to retrieve these values.
- Fixed multiple issues where FLR from multiple VFs are not completing.
- Added new RSS hash secret key modes.
- Fixed an issue where LVDS output was not getting enabled using vpd.
DCBX:
- Fixed an issue where iscsi tlv is sent incorrectly to host (DCBX CEE).
- Fixed an issue where app priority values are not handled correctly
in fw (DCBX IEEE).
ETH:
- Adjusts the link related delay timings according to the QSFP spec.
- Changed 2.5G mac speed bit to 25G mac speed bit in fw API.
- Improvement in 40G link bringup time with few switches.
OFLD:
- Do not reserve qp/cq if rdma capability is not enabled.
- Fixed an issue where approx 1600+ TOE connections were causing a
firmware fatal error.
- Fixed DSGL memory write in T5. Now iwarp and iscsi can use DSGL to do
memory write.
- Fixed multiple issues in hash filter mode where incorrect protocol
mask was getting used and affecting hash filter functionality.
- New fastpath WR FW_RI_FR_NSMR_TPTE_WR (with fully populated TPTE) is
added for small REG_MR operations.
FOiSCSI:
- Fixed an issue in foiscsi recovery path.
- Fixed an issue where foiscsi (in VM in PCIE passthrough mode) didn't
come up after VM FLR.
ENHANCEMENTS
------------
ETH:
- Implemented 1G/10G KR/KX ability.
- Implemented T540-BT adapter support.
=======================================================================
T4 Firmware
+++++++++++
Version : 1.16.12.0
Date : 11/11/2016
FIXES
-----
BASE:
- Fixed an issue where reading temperature sesors using ldst command
causes mailbox timeout.
- Added new FW_PARAMS_CMD[DEV] options to retrieve Serial Configuration
and VPD version numbers.
ETH:
- Fixed DCBX CEE Incorrect class to pririty mapping.
FOiSCSI:
- Fixed an issue where unloading foiscsi driver causes mailbox timeout.
MFC after: 3 days
Sponsored by: Chelsio Communications
Instead of spuriously re-reading the lock value, read it once.
This change also has a side effect of fixing a performance bug:
on failed _mtx_obtain_lock, it was possible that re-read would find
the lock is unowned, but in this case the primitive would make a trip
through turnstile code.
This is diff reduction to a variant which uses atomic_fcmpset.
Discussed with: jhb (previous version)
Tested by: pho (previous version)
It was a temporary change to ease an import of native atomic_cas primitives.
Instead, atomic_fcmpset was devised with different semantics. See r311168.
The micro-Firewall DTS is now a single BSD licensed file.
Reported by: manu
Obtained from: pfSense
MFC after: 2 weeks
Sponsored by: Rubicon Communications, LLC (Netgate)
Original log:
Do not initialize the adapter on MTU change when adapter status is down.
This fixes long-standing problems when changing settings of the adapter.
Discussed in:
https://lists.freebsd.org/pipermail/freebsd-net/2016-June/045509.html
Reported by: Franco Fichtner <franco@opnsense.org>
MFH: 2 days
On systems without a configured swap device, an attempt to launder pages
from a swap object will always fail and result in the page being
reactivated. This means that the page daemon will continuously scan pages
that can never be evicted. With this change, anonymous pages are instead
moved to PQ_UNSWAPPABLE after a failed laundering attempt when no swap
devices are configured. PQ_UNSWAPPABLE is not scanned unless a swap device
is configured, so unreferenced unswappable pages are excluded from the page
daemon's workload.
Reviewed by: alc
and prison enforcement. Do it on the caller buffer directly.
Besides eliminating memory copies, this change also removes large
structure from the kernel stack.
Extracted from: ino64 work by gleb
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
- iflib - add checksum in place support (mmacy)
- iflib - initialize IP for TSO (going to be needed for e1000) (mmacy)
- iflib - move isc_txrx from shared context to softc context (mmacy)
- iflib - Normalize checks in TXQ drainage. (shurd)
- iflib - Fix queue capping checks (mmacy)
- iflib - Fix invalid assert, em can need 2 sentinels (mmacy)
- iflib - let the driver determine what capabilities are set and what
tx csum flags are used (mmacy)
- add INVARIANTS debugging hooks to gtaskqueue enqueue (mmacy)
- update bnxt(4) to support the changes to iflib (shurd)
Some other various, sundry updates. Slightly more verbose changelog:
Submitted by: mmacy@nextbsd.org
Reviewed by: shurd
mFC after:
Sponsored by: LimeLight Networks and Dell EMC Isilon
Add two new qualifiers for use by the static checkers:
_Nonnull
The _Nonnull nullability qualifier indicates that null is not a meaningful
value for a value of the _Nonnull pointer type.
_Nullable
The _Nullable nullability qualifier indicates that a value of the
_Nullable pointer type can be null.
These were introduced in Clang 3.7. For more information, see:
http://clang.llvm.org/docs/AttributeReference.html#nonnull
We add these now without using them so that the GCC ports have time to
pick up the header change.
Hinted by: Android Bionic libc [1]
Also seen in: Apple's Libc-1158.20.4
[1]
baa2a973bd
(disabled by default).
To enable it, dev.rtwn.%d.radar_detection tunable need to be set
to a nonzero value.
Tested with RTL8821AU, AP mode (no radar events were seen - so,
I have no idea if it is really correct / working)
Reviewed by: adrian
Differential Revision: https://reviews.freebsd.org/D8903
structs under the INET6 #ifdef. Similarly (even though it doesn't seem
to affect the build), conditionalize all IPv4 structs under the INET
#ifdef
This also unbreaks the LINT-NOINET6 tinderbox target on amd64; I have not
verified other MACHINE/TARGET pairs (e.g. armv6/arm).
MFC after: 2 weeks
X-MFC with: r310847
Pointyhat to: jpaetzel
Reported by: O. Hartmann <o.hartmann@walstatt.org>
Apparently the libstand dosfs optimization is a bit too optimistic
and did introduce possible memory corruption.
This patch is backing out the bad part and since this results in
dosfs reading full blocks now, we can also remove extra offset argument
from dv_strategy callback.
The analysis of the issue and the backout patch is provided by Mikhail Kupchik.
PR: 214423
Submitted by: Mikhail Kupchik
Reported by: Mikhail Kupchik
Reviewed by: bapt, allanjude
Approved by: allanjude (mentor)
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D8644
If pager' populate method succeeded, but other thread raced with us
and modified vm_map, we must unbusy all pages busied by the pager,
before we retry the whole fault handling. If pager instantiated more
pages than fit into the current map entry, we must unbusy the pages
which are clipped.
Also do some refactoring, clarify comments and use more clear local
variable names.
Reported and tested by: kargl, subbsd@gmail.com (previous version)
Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
If there is a loop in the network a CARP that is in MASTER state will see it's
own broadcasts, which will then cause it to assume BACKUP state. When it
assumes BACKUP it will stop sending advertisements. In that state it will no
longer see advertisements and will assume MASTER...
We can't catch all the cases where we are seeing our own CARP broadcast, but
we can catch the obvious case.
Submitted by: torek
Obtained from: FreeNAS
MFC after: 2 weeks
Sponsored by: iXsystems
The boot2 family of bootblocks (zfsboot/gptzfsboot) are using separate
implementation if keyboard reading code, which has deadlock case when
extended key (arrows etc) are pressed.
The problem is about avoiding the noise from some systems, generating
false key events with scan code 1 and ascii code 00, so the code
does attempt to filter such cases out. Unfortunately the extended keys
also set ascii 0, and therefore the pressed key event is ignored and
the keypress is never read, resulting in infinite loop.
This update is moving the check to keyhit() function and is allowing
the rest of the code to process the extended keys.
Reviewed by: bapt, allanjude
Approved by: allanjude (mentor)
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D8608
* Firmware oriented NICs may need to sleep in their configuration paths.
Use RLOCK instead of WLOCK to allow this to again occur.
This fixes netmap on cxgbe.
* Change the worker lock to a normal mutex rather than a spin lock.
Drivers shouldn't be doing netmap work from the fast interrupt
handlers, so it's not required to be a spinlock.
Submitted by: luigi, Vincenzo Maffione <v.maffione@gmail.com>
Reviewed by: jhb
This allows the common code to use the correct type for the lock state
local variable passed to EFSYS_LOCK() and EFSYS_UNLOCK().
On Windows, this allows warning supression pragmas to be removed.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Wake-on-lan is not supported in production on any of our adapters, as
they don't have the required AUX power connector. (It's possible that
AUX power is supplied to some of our ALOM or mezz adapters, but if so
then we've never implemented or tested WoL support.)
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8972
The previous API had various problems, including the length of the
caller provided buffer not being specified, no means being available
to discover how big the buffer needs to be, and a lack of clarity of
what the resulting list contains.
To fix it:
- add the buffer length as a parameter
- if the provided buffer is too short, fail with ENOSPC and return the
required length
- ensure that the list contents are valid and add comments
describing it
It is safe to change this API as, unsuprisingly, it has no users.
Submitted by: Mark Spender <mspender at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8971
Manftest needs to know exactly what went wrong with the verified update
so that failing boards can be correctly diagnosed.
Submitted by: Tom Millington <tmillington at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8969
In newer firmware that supports multithreaded MCDI processing,
longer running commands may be run ina background thread. Add
support for drivers to query the appropriate timeout for each
MCDI request.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8968
Poll-mode driver does not use interrupts and number of used event queues
should not be limitted by the number of interrupts allocated for the
NIC.
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8967
- Update struct link_settings and associated shared code.
- Add tunables to control FEC and autonegotiation. All ports inherit
these values as their initial settings.
hw.cxgbe.fec
hw.cxgbe.autoneg
- Add per-port sysctls to control FEC and autonegotiation. These can be
modified at any time.
dev.<port>.<n>.fec
dev.<port>.<n>.autoneg
MFC after: 3 days
Sponsored by: Chelsio Communications
This patch solves IRQ generation problems using the mlx5en(4) driver
with xenserver v6.5.0 in SRIOV and PCI-passthrough modes.
Until further the hw.pci.msix_rewrite_table quirk must be set manually
in /boot/loader.conf .
Reviewed by: jhb @
Sponsored by: Mellanox Technologies
MFC after: 2 weeks
Recognize new MACHINE_ARCH names now as we have added hardfloat support.
Switch JZ4780 to mipselhf and remove all uses of TARGET_ARCH in kernel
.mk files.
Reviewed by: adrian
Differential Revision: https://reviews.freebsd.org/D8989
QEMU does not implement hardware debug registers so when
dbg_monitor_is_enabled is called kernel receives "invalid instruction"
exception. QEMU implements only DIDR register and on read returns all
zeroes to indicate that it doesn't support other registers. Real
hardware has Version bits set.
MFC after: 1 week
Ingenic CPUs treat plain cache writeback as local-only operation and do
nothing if that is a remote CPU that holds the dirty cache line. They
do broadcast invalidate and write-and-invalidate to other cores though,
so take advantage of that and use wbinv in place of wb as this still gives
us required busdma semantics. Otherwise we'd have to do IPI to remote CPU
ourselves.
- Since I/Os are allocates from per-port pools, make allocations store
pointer to CTL softc there, and use it where needed instead of global.
- Created bunch of helper macros to access LUN, port and CTL softc.
MFC after: 2 weeks
transmitter, but not both at the same time. This patch:
- Adds a dev.pcm.0.internal_codec sysctl node for selecting between
internal and external codec
- Changes playback sample rate from 96 kHz to 48 kHz for HDMI compatibility
- Enables i2s clock on codec access
Reviewed by: br
Differential Revision: https://reviews.freebsd.org/D8960
All hash sizes are power-of-2, but the compiler does not know that for sure
and 'foo % size' forces doing a division.
Store the size - 1 and use 'foo & hash' instead which allows mere shift.
ICC complains that enumerated type mixed with another type.
Found by DPDK upstream build sanity check.
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8954
It is not 100% correct to assign non-enum values to enum type
variables.
Found by ICC build (DPDK PMD upstreaming).
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8953
It is not safe to push TSO option descriptors if pacer bypass is
enabled, so to make sure that doesn't happen never push TSO option
descriptors.
Submitted by: Mark Spender <mspender at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8952
The flag EFX_FILTER_MATCH_LOC_MAC_IG to represent filtering on the
individual/group bit of the MAC address (with the two cases being
distingusished by the MAC address in the filter specification) was
introduced to mirror the Linux driver filtering code, but the
implementations are different enough anyway that it isn't of much value.
Having separate flags for unknown unicast and multicast simplifies
the code and allows the set of flags to match those used by MCDI.
It will also makes it easier to report whether these filters are
supported.
In the MCDI definitions, the unknown multicast and unicast flags have
the values 0x40000000 and 0x80000000 respectively, and so using the
same values for simplicity requires 32 bits in the filter specification
to store the flags. This means the structure is now a little bigger
than 64 bytes, but filters are not often used on critical paths so this
shouldn't have much impact - on Linux they are also bigger than they
used to be.
Submitted by: Mark Spender <mspender at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8951
It is required to minimize RxQ context in the driver or avoid chaising
for the NIC handle in adapter (global per-interface) structure.
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8950
The semantics of the MCDI interfacve require reading the first
dword of the header before any other data in the buffer. Add
a barrier to the common code MCDI handler to enforce this.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8949
The port mask used for per-port sensors in mcdi_sensor_map
assumes zero-based port numbering. The port mask used in
the code is based on the one-based MCDI port number.
Fix this to lookup the correct per-port sensors, and to
allow reporting of sensor events from higher port numbers.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8947
Both Siena and EF10 use the siena_ev_qpoll() implementation, but this
function is not defined in builds without EFSYS_OPT_SIENA.
Remove siena_ev_qpoll and inline it into efx_ev_qpoll to allow it
to be used in non-Siena builds.
Also remove outdated FIXME comment, as EF10 event batching/merging has
been implemented long ago without needing to modify this code.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8939
Clang disallows ambiguous instructions (GNU as has a default based on
chosen .code setting). We only need 'orb' here because KARGS_FLAGS_PXE
fits in a byte; this is the same as done in bxeboot.
Reviewed by: kib
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D8959
add support for object types that were previously prohibited because they
could contain PG_CACHED pages.
Roughly halve the number of radix trie operations performed by
vm_page_alloc_contig() using the same approach that is employed by
vm_page_alloc(). Also, eliminate the radix trie lookup performed with the
free page queues lock held.
Tidy up the handling of radix trie insert failures in vm_page_alloc() and
vm_page_alloc_contig().
Reviewed by: kib, markj
Tested by: pho
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D8878
The function flags were changed to mirror the privileges, but
the privileges are preferred.
Submitted by: Mark Spender <mspender at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8936
Submitted by: Andrew Lee <alee at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8934
On Medford, using MC_CMD_GET_RXDP_CONFIG to query the RX end
padding setting is in the ADMIN group, and so fails for
unprivileged functions. In that case, assume the largest size
supported by Medford hardware (256bytes) to prevent overrun.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8933
For Siena and Huntington, the per-port bootcfg (aka expcfg) is
stored in a dedicated 4Kbyte partition for each port.
For Medford, the per-PF bootcfg is stored in a 2Kbyte sector
within a single shared partition. Update the common code to support
the new bootcfg layout.
Submitted by: Andy Moreton <amoreton at solarflare.com>
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8932
It is ignored on SFN5xxx/6xxx (aka Siena).
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8931
This is needed because MCDI command MC_CMD_REKEY can return
MC_CMD_ERR_ERANGE.
Submitted by: Tom Millington <tmillington at solarflare.com>
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
With default settings GNU ld generates two PT_LOADs for loader.sym while
LLD generates three, because it creates a rodata segment. Previously
btxldr terminated phdr processing after two PT_LOADs. Remove the early
termination to process all PT_LOADs.
Reviewed by: kib, tsoome
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D8929
Some MIPS revisions do implement uncached-accelerate caching
attribute, but place extra requirement on access, such as
partial-word or out-of-sequence writes potentially having an
“unpredictable” effects.
announced by HDA controller.
Incorrectly implermented HDA codec may report support for more stripes
that HDA controller already have. Due to this, always limit number of
enabled stripes by global controller maximum.
Reviewed by: mav
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D8922
It is simpler and cleaner to have strip produce the stripped output
directly than copying the input to the output first.
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
On platforms that have uncached-accelerate cache attribute, map it
to VM_MEMATTR_WRITE_COMBINING. Otherwise, leave write comining
undefined.
Reviewed by: adrian, jhb (glance)
Differential Revision: https://reviews.freebsd.org/D8894
We do not have non-volatile memory to really save those values, so we
neither report nor support this capability. Also saved mode pages are
not replicated between HA peers now.
MFC after: 2 weeks
Add SPI mode (PIO-only) support for Intel Synchronous Serial Port that
can be found in several Intel's products starting from PXA family.
Most of implementations have slight differences in behavior and in
addresses for registers subset. This driver covers only BayTrail SoC
implementation for it's the only hardware I have to test it on.
Driver attaches to ACPI bus only and does not have PCI or FDT support
for now due to lack of hardware to test it on.
"intelspi" is the best name I've managed to come up with. Linux driver
name (spi-pxa2xx) does not make sense because current implementation
does not support actual PXA2xx SoCs. And as far as I know there is no
codename assigned to Intel SSP chip.
Reviewed by: br, manu
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D8896
This argument is not a bitmask of flags, but only accepts a single value.
Fail with EINVAL if an invalid value is passed to 'flag'. Rename the
'flags' argument to getmntinfo(3) to 'mode' as well to match.
This is a followup to r308088.
Reviewed by: kib
MFC after: 1 month
Rework general approach to locking and working with audio worker thread:
- Use flags to signal requested worker action
- Fix submitted buffer calculations to avoid samples starvation
- Protect buffer pointers with locks to fix race condition between callback
and audio worker thread
- Remove unnecessary vchi_service_use
- Do not use lock to serialize VCHI requests since only one thread issues them now
- Fix unloading signaling per hselasky@ suggestion
- Add output to detect inconsistent callback data caused by possible firmware bug
https://github.com/raspberrypi/firmware/issues/696
- Add stats/debug sysctls to troubleshoot possible bugs
PR: 213687, 205979, 215194
MFC after: 1 week
Those two values are not directly related, so make them independent.
This does not change any limits immediately, but makes number of LUNs
per port controllable via tunable/sysctl kern.cam.ctl.lun_map_size.
After this change increasing CTL_MAX_LUNS should be pretty cheap,
and even making it tunable should be easy.
MFC after: 2 weeks
Specification of entire RSS table in the driver allows to spread traffic
more equally across CPUs/RSS channels if number of RSS channels is not
power of 2.
Reviewed by: philip
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D8910
The ath10k firmware supports hardware WEP offload, and in native wifi mode
(or 802.3 ethernet mode, for that matter) the WEP key isn't actually included
in the TX payload from net80211. Instead, a separate firmware command is issued
that sets the default TX key to be the specified key.
However, net80211 doesn't at all inform the driver layer that this is
occuring - it just "expects" to be inserting WEP header information
when doing WEP TX, even with hardware encryption.
So, to better support the newer world order, turn the default TX key assignment
into a VAP method that can be overridden by the driver and ensure its wrapped
in a crypto begin/end set. That way it should be correctly atomic from the
point of view of keychanges (as long as the driver does the right thing.)
It'd be nice if we passed through to the key_set call a flag that says
"also make this the default key" - that's captured here by calling the
deftxkey method after the key_set method. Maybe I can do that later.
Note: this is a net80211 ABI change, and will require a kernel+modules
recompile. Happy Holidays, etc.
Tested:
* ath10k driver port
* rtwn_usb, WEP station
Add ACPI part for ig4 driver to make it work on Intel BayTrail SoC where
ig4 device is available only through ACPI
Reviewed by: avg
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D8742
closed by r310302 for knote().
If KN_INFLUX | KN_SCAN flags are set for the note passed to knote() or
knote_fork(), i.e. the knote is scanned, we might erronously clear
INFLUX when finishing notification. For normal knote() it was fixed
in r310302 simply by remembering the fact that we do not own
KN_INFLUX, since there we own knlist lock and scan thread cannot clear
KN_INFLUX until we drop the lock. For knote_fork(), the situation is
more complicated, e must drop knlist lock AKA the process lock, since
we need to register new knotes.
Change KN_INFLUX into counter and allow shared ownership of the
in-flux state between scan and knote_fork() or knote(). Both in-flux
setters need to ensure that knote is not dropped in parallel. Added
assert about kn_influx == 1 in knote_drop() verifies that in-flux state
is not shared when knote is destroyed.
Since KBI of the struct knote is changed by addition of the int
kn_influx field, reorder kn_hook and kn_hookid to fill pad on LP64
arches [1]. This keeps sizeof(struct knote) to same 128 bytes as it
was before addition of kn_influx, on amd64.
Reviewed by: markj
Suggested by: markj [1]
Tested by: pho (previous version)
Sponsored by: The FreeBSD Foundation
Differential revision: https://reviews.freebsd.org/D8898
There are two instances of inlined unlocks + continue in vmtotal()
switch statements, which are ordinary expressed with break from the
switch case and code after the switch. Also, the combination of
continue and break statement is redundand.
Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
accepting the wrong state and printing warning. Do not obliterate
kl_lock and kl_unlock pointers, they are often useful for post-mortem
analysis.
Reviewed by: markj
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
X-Differential revision: https://reviews.freebsd.org/D8898
There was report of stable/11 build failure on older versions of
FreeBSD. The problem was a reference to static variable in inline
function. Although fairly recent versions of clang can cope with this
situation just fine there is no strict requirement to make functions
in question inline, they're not in any time-critical codepath. So to
keep HEAD and 11-STABLE buildable on older FreeBSDs some of the util
functions were converted from inline to normal.
Reported by: hselasky
MFC after: 3 days
and host1x module. Unfortunately, tegra124 SoC doesn't have 2D acceleration
engine and 3D requires not yet started nouveau driver.
These drivers forms a first non-x86 DRM2 enabled graphic stack.
Note, there are 2 outstanding issues:
- The code uses gross hack in order to be comply with
OBJT_MGTDEVICE pager. (See tegra_bo_init_pager() in tegra_bo.c)
- Due to improper(probably) refcounting in drm_gem_mmap_single()
(in drm_gem.c), the gem objects are never released.
I hope that I will be able to address both issues in finite time,
but I don't want to touch x86 world now.
MFC after: 1 month
If system starts without attached monitor, DRM create framebuffer
for VT console. Later, when monitor is attached, the hotplug event must
issue full modeset procedure to setup CRTC.
In original code, this was done in drm_fb_helper_set_par(), but we don't
have this function implemented yet. Use unrolled version of
drm_fb_helper_set_par() to ensure same functionality.
MFC after: 1 month
Although vchiq_build_date and vchiq_build_time are not used in current
vchi driver at the moment, make sure these value will not leak into
build later on if at some point they will be refered in some new
imported code
PR: 215494
Reported by: emaste
MFC after: 1 week
There is no need to do two allocations per kqueue timer. Gather all
data needed by the timer callout into the structure and allocate it at
once.
Use the structure to preserve the result of timer2sbintime(), to not
perform repeated 64bit calculations in callout.
Remove tautological casts.
Remove now unused p_nexttime [1].
Noted by: markj [1]
Reviewed by: markj (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
X-MFC note: do not remove p_nexttime
Differential revision: https://reviews.freebsd.org/D8901
the ones obtained through devclass_get_device(). Some minor code
cleanups while at it.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
In case of the empty queue tp->snd_holes and tcp_sackhole_insert()
failing due to memory shortage, tp->snd_holes will be empty.
This problem was hit when stress tests where performed by pho.
PR: 215513
Reported by: pho
Tested by: pho
Sponsored by: Netflix, Inc.
For EXTENDED COPY:
- improve parameters checking to report some errors before copy start;
- forward sense data from copy target as descriptor in case of error;
- report which CSCD reported error in sense key specific information.
For WRITE USING TOKEN:
- pass through real sense data from copy target instead of reporting
our copy error, since for initiator its a "simple" write, not a copy.
MFC after: 2 weeks
- Allow maximal sense size limitation via Control Extension mode page.
- When sense size limited, include descriptors atomically: whole or none.
- Set new SDAT_OVFL bit if some descriptors don't fit the limit.
- Report real written sense length instead of static maximal 252 bytes.
MFC after: 2 weeks
The count argument natural type if vm_pindex_t, but due to the loop
organization, it has to be signed type to detect the termination
condition. Replace this logic by using distinguished counter for the
processed pages, and terminate loop when the counter exceeds the
argument.
Completely process one swblock for all relevant indexes instead of
doing relookup in hash when incrementing page index on the loop step.
Do not drop hash mutex around iterations.
Noted and reviewed by: alc
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Return BUS_PROBE_NOWILDCARD in probe method to make sure that spigen
attaches only to the device created in identify method.
Before this change spigen probe method used to return 0 which meant it
competed with other drivers to be attached to the devices created for
child nodes of SPI bus node in FDT.
Reported by: Daniel Braniss
MFC after: 1 week
For most NFSv4.1 servers, a NFS4ERR_BAD_SESSION error is a rare failure
that indicates that the server has lost session/open/lock state.
However, recent testing by cperciva@ against the AmazonEFS server found
several problems with client recovery from this due to it generating this
failure frequently.
Briefly, the problems fixed are:
- If all session slots were in use at the time of the failure, some processes
would continue to loop waiting for a slot on the old session forever.
- If an RPC that doesn't use open/lock state failed with NFS4ERR_BAD_SESSION,
it would fail the RPC/syscall instead of initiating recovery and then
looping to retry the RPC.
- If a successful reply to an RPC for an old session wasn't processed
until after a new session was created for a NFS4ERR_BAD_SESSION error,
it would erroneously update the new session and corrupt it.
- The use of the first element of the session list in the nfs mount
structure (which is always the current metadata session) was slightly
racey. With changes for the above problems it became more racey, so all
uses of this head pointer was wrapped with a NFSLOCKMNT()/NFSUNLOCKMNT().
- Although the kernel malloc() usually allocates more bytes than requested
and, as such, this wouldn't have caused problems, the allocation of a
session structure was 1 byte smaller than it should have been.
(Null termination byte for the string not included in byte count.)
There are probably still problems with a pNFS data server that fails
with NFS4ERR_BAD_SESSION, but I have no server that does this to test
against (the AmazonEFS server doesn't do pNFS), so I can't fix these yet.
Although this patch is fairly large, it should only affect the handling
of NFS4ERR_BAD_SESSION error replies from an NFSv4.1 server.
Thanks go to cperciva@ for the extension testing he did to help isolate/fix
these problems.
Reported by: cperciva
Tested by: cperciva
MFC after: 3 months
Differential Revision: https://reviews.freebsd.org/D8745
operations to the MD PCPU region. Change sysmap initialization to only
allocate KVA pages for CPUs that are actually present. As a minor
optimization, this also prevents false sharing between adjacent sysmap objects
since the pcpu struct is already cacheline-aligned.
While here, move pc_qmap_addr initialization for the BSP into
pmap_bootstrap(), which allows use of pmap_quick* functions during early boot.
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D8833
It is not compat w/ the old timesync message format, which the message
type stays the same as the old timesync message.
MFC after: 1 week
Sponsored by: Microsoft
Kernel stack overflows in MIPS call panic() directly from an assembly
handler after storing the interrupted context's registers in a
trapframe. Rather than inferring the location of ra, sp, and pc from
the instruction stream, recognize the pc of a kernel stack overflow
and pull the registers from the trapframe.
Sponsored by: DARPA / AFRL
- Honor PG_NODUMP by not dumping pages with this flag set.
- Pat the watchdog during dumps to avoid a watchdog reset while writing
out a dump.
- Reformat the output during a dump to update every 10% done rather than
every 2MB dumped.
- Include UMA small pages and pages holding PV entries in minidumps.
Sponsored by: DARPA / AFRL
to recreate it from ${MAKEOBJDIRPREFIX} and ${SRC_BASE} and ${KERNCONF},
the latter being especially problematic when KERNCONF is set to the names
of multiple kernel configs.
I don't yet know which RX descriptor bits map to shortgi, long-gi,
short-preamble, long-preamble, STBC, LDPC, HT40, etc - so I can't
easily add those just yet.
There's apparently no per-frame RX RSSI information exposed so we
also just use the results from the previous calibration task.
This also tidies up how the per-mbuf RSSI is pushed into the frame -
now that it's attached to the mbuf via rx_stats, we don't have to
do any silly hijinx to get it out of the frame processing path.
Tested:
* RTL8712, 1x1 cut 3, STA mode
dump_avail[] is supposed to be a superset of phys_avail[] that
describes all of the memory ranges that should be included in a full
dump. minidumps don't consider pages described by dump_avail[] to be
valid and thus they are excluded via the is_dumpable() function. Most
MIPS platforms (including MALTA) set dump_avail[] to be identical to
phys_avail[]. In particular, phys_avail[] doesn't include the kernel
itself, so pages for the kernel and it's global variables are not
considered dumpable and not included in the dump. Fix this by setting
dump_avail[0] to the first memory address (0) rather than the end of
the kernel.
Several other MIPS platforms have the same bug, though I am only able
to test malta in qemu. The correct fix is to set dump_avail[] to
describe RAM and in particular to not set dump_avail[0] to the end of
the kernel (kernel_kseg0_end).
Sponsored by: DARPA / AFRL
The removal of TAILQ_FOREACH_SAFE introduced a small race: when the last
thread on a sleepqueue is awoken, it reclaims the sleepqueue and may begin
executing on a different CPU before sleepq_resume_thread() returns. This
leaves a window during which it may go back to sleep and incorrectly be
awoken again by the caller of sleepq_broadcast().
Reported and tested by: pho
MFC after: 3 days
Sponsored by: Dell EMC Isilon
Current Xen IPI setup functions require that the caller provide a device in
order to obtain the name of the interrupt from it. With early AP startup this
device is no longer available at the point where IPIs are bound, and a KASSERT
would trigger:
panic: NULL pcpu device_t
cpuid = 0
KDB: stack backtrace:
db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xffffffff82233a20
vpanic() at vpanic+0x186/frame 0xffffffff82233aa0
kassert_panic() at kassert_panic+0x126/frame 0xffffffff82233b10
xen_setup_cpus() at xen_setup_cpus+0x5b/frame 0xffffffff82233b50
mi_startup() at mi_startup+0x118/frame 0xffffffff82233b70
btext() at btext+0x2c
Fix this by no longer requiring the presence of a device in order to bind IPIs,
and simply use the "cpuX" format where X is the CPU identifier in order to
describe the interrupt.
Reported by: sbruno, cperciva
Tested by: sbruno
X-MFC-With: r310177
Sponsored by: Citrix Systems R&D
