Add an ocxl driver to handle generic opencapi devices. Of course, it's
not meant to be the only opencapi driver, any device is free to
implement its own. But if a host application only needs basic services
like attaching to an opencapi adapter, have translation faults handled
or allocate AFU interrupts, it should suffice.
The AFU config space must follow the opencapi specification and use
the expected vendor/device ID to be seen by the generic driver.
The driver exposes the device AFUs as a char device in /dev/ocxl/
Note that the driver currently doesn't handle memory attached to the
opencapi device.
Signed-off-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Signed-off-by: Alastair D'Silva <alastair@d-silva.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Jeff Kirsher says:
====================
10GbE Intel Wired LAN Driver Updates 2018-01-23
This series contains updates to ixgbe only.
Shannon Nelson provides an implementation of the ipsec hardware offload
feature for the ixgbe driver for these devices: x540, x550, 82599.
The ixgbe NICs support ipsec offload for 1024 Rx and 1024 Tx Security
Associations (SAs), using up to 128 inbound IP addresses, and using the
rfc4106(gcm(aes)) encryption. This code does not yet support checksum
offload, or TSO in conjunction with the ipsec offload - those will be
added in the future.
This code shows improvements in both packet throughput and CPU utilization.
For example, here are some quicky numbers that show the magnitude of the
performance gain on a single run of "iperf -c <dest>" with the ipsec
offload on both ends of a point-to-point connection:
9.4 Gbps - normal case
7.6 Gbps - ipsec with offload
343 Mbps - ipsec no offload
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The Atomic Operations feature (PCIe r4.0, sec 6.15) allows atomic
transctions to be requested by, routed through and completed by PCIe
components. Routing and completion do not require software support.
Component support for each is detectable via the DEVCAP2 register.
A Requester may use AtomicOps only if its PCI_EXP_DEVCTL2_ATOMIC_REQ is
set. This should be set only if the Completer and all intermediate routing
elements support AtomicOps.
A concrete example is the AMD Fiji-class GPU (which is capable of making
AtomicOp requests), below a PLX 8747 switch (advertising AtomicOp routing)
with a Haswell host bridge (advertising AtomicOp completion support).
Add pci_enable_atomic_ops_to_root() for per-device control over AtomicOp
requests. This checks to be sure the Root Port supports completion of the
desired AtomicOp sizes and the path to the Root Port supports routing the
AtomicOps.
Signed-off-by: Jay Cornwall <Jay.Cornwall@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[bhelgaas: changelog, comments, whitespace]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Some multifunction PCI devices with more than 8 functions use "alternative
routing-ID interpretation" (ARI), which means the 8-bit device/function
number field will be interpreted as 8 bits specifying the function number
(the device number is 0 implicitly), rather than the upper 5 bits
specifying the device number and the lower 3 bits specifying the function
number. The kernel can enable and use this.
Expose in a sysfs attribute whether the kernel has enabled ARI, so that a
program in userspace won't have to parse PCI devices and PCI configuration
space to figure out if it is enabled. This will allow better predictable
network naming using PCI function numbers without using PCI bus or device
numbers, which is desirable because bus and device numbers can change with
system configuration but function numbers will not.
Signed-off-by: Stuart Hayes <stuart.w.hayes@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Certain Thunderbolt 1 controllers claim to support Command Completed events
(value of 0b in the No Command Completed Support field of the Slot
Capabilities register) but in reality they neither set the Command
Completed bit in the Slot Status register nor signal a Command Completed
interrupt:
8086:1513 CV82524 [Light Ridge 4C 2010]
8086:151a DSL2310 [Eagle Ridge 2C 2011]
8086:151b CVL2510 [Light Peak 2C 2010]
8086:1547 DSL3510 [Cactus Ridge 4C 2012]
8086:1548 DSL3310 [Cactus Ridge 2C 2012]
8086:1549 DSL2210 [Port Ridge 1C 2011]
All known newer chips (Redwood Ridge and onwards) set No Command Completed
Support, indicating that they do not support Command Completed events.
The user-visible impact is that after unplugging such a device, 2 seconds
elapse until pciehp is unbound. That's because on ->remove,
pcie_write_cmd() is called via pcie_disable_notification() and every call
to pcie_write_cmd() takes 2 seconds (1 second for each invocation of
pcie_wait_cmd()):
[ 337.942727] pciehp 0000:0a:00.0:pcie204: Timeout on hotplug command 0x1038 (issued 21176 msec ago)
[ 340.014735] pciehp 0000:0a:00.0:pcie204: Timeout on hotplug command 0x0000 (issued 2072 msec ago)
That by itself has always been unpleasant, but the situation has become
worse with commit cc27b735ad ("PCI/portdrv: Turn off PCIe services during
shutdown"): Now pciehp is unbound on ->shutdown. Because Thunderbolt
controllers typically have 4 hotplug ports, every reboot and shutdown is
now delayed by 8 seconds, plus another 2 seconds for every attached
Thunderbolt 1 device.
Thunderbolt hotplug slots are not physical slots that one inserts cards
into, but rather logical hotplug slots implemented in silicon. Devices
appear beyond those logical slots once a PCI tunnel is established on top
of the Thunderbolt Converged I/O switch. One would expect commands written
to the Slot Control register to be executed immediately by the silicon, so
for simplicity we always assume NoCompl+ for Thunderbolt ports.
Fixes: cc27b735ad ("PCI/portdrv: Turn off PCIe services during shutdown")
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: stable@vger.kernel.org # v4.12+
Cc: Sinan Kaya <okaya@codeaurora.org>
Cc: Yehezkel Bernat <yehezkel.bernat@intel.com>
Cc: Michael Jamet <michael.jamet@intel.com>
Cc: Andreas Noever <andreas.noever@gmail.com>
Fix recreating the channel VSIs during the reset flow to reconfigure
the Tx rings and the queue context associated with the channel VSI.
Also update the next_base_queue for the VSI while rebuilding the
channel VSIs after a reset.
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
If we receive the link status message from PF with link up before queues
are actually enabled, it will trigger a TX hang. This fixes the issue
by ignoring a link up message if the VF state is not yet in RUNNING
state.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Client close is overloaded to handle both un-registration and
netdev down event. On netdev down, i40iw client close is called
which unregisters the RDMA dev and this is too destructive
since the netdev is still registered.
Do not call client close/open on netdev down/up events. Instead
disable the PE TCP_ENA flag during a netdev down event. This
blocks all TCP traffic to the RDMA Protocol Engine. On netdev up,
re-enable the flag.
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The driver (and the entire netdev layer for that matter) assumes
that TC0 will always be present in our DCB configuration.
Unfortunately, this isn't always the case. Rather than fail to
configure the VSI, let's go ahead and try to make it work, even
though DCB will end up being disabled by the kernel.
If the driver fails to configure DCB, the driver queries what's
valid, then writes that back to the hardware, always forcing TC0.
This fixes a bug where the driver could fail to adhere to ETS BW
allocations if 8 TCs were configured on the switch.
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
In VFs, there is a known issue which can cause writebacks
to not occur when interrupts are disabled and there are
less than 4 descriptors resulting in TX timeout. Timeout
can also occur due to lost interrupt.
The current implementation for detecting and recovering
from hung queues in the PF is problematic because it actually
actively encourages lost interrupts. By triggering a SW
interrupt, interrupts are forced on. If we are already in
napi_poll and an interrupt fires, napi_poll will not be
rescheduled and the interrupt is effectively lost; thereby
potentially *causing* hung queues.
This patch checks whether packets are being processed between
every watchdog cycle and determine potential hung queue and
fires triggers SW interrupt only for that particular queue.
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This fix solves an issue occurring while calling i40e_led_set function
from the driver with "blink" parameter set as TRUE. This call resulted
in Activity LED blinking instead of Link LED, which may lead to errors
in physically identifying the port, since Activity LED may be blinking
for different reasons as well.
Signed-off-by: Michal Kuchta <michal.kuchta@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When a host disables and enables a PF device, all the associated
VFs are removed and added back in. It also generates a PFR which in turn
resets all the connected VFs. This behaviour is different from that of
Linux guest on Linux host. Hence we end up in a situation where there's
a PFR and device removal at the same time. And watchdog doesn't have a
clue about this and schedules a reset_task. This patch adds code to send
signal to reset_task that the device is currently being removed.
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
flush_schedule_work blocks until completion of all scheduled
work items in global work-queue. This can cause deadlock in some
cases. i40evf_remove() cleans up necessary work items with
cancel_delayed_work_sync and cancel_work_sync. This fix removes
flush_schedule_work call inside i40evf_remove().
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
In some weird circumstances with DCB enabled, the firmware can fail to
configure the VSI, leaving us with zero traffic classes. Check for this
state when we configure RSS to avoid a panic.
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds new I40E_NVMUPD_GET_AQ_EVENT state to allow
retrieval of AdminQ events as a result of AdminQ commands sent
to firmware.
Add preservation flags support on X722 devices for NVM update
AdminQ function wrapper. Add new parameter and handling to
nvmupdate admin queue function intended to allow nvmupdate tool
to configure the preservation flags in the AdminQ command.
This is required to implement FlatNVM on X722 devices.
Signed-off-by: Pawel Jablonski <pawel.jablonski@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The queue count says the highest queue that's been allocated, so don't
reallocate a queue lower than that.
Fixes: 147b27e4bd ("nvme-pci: allocate device queues storage space at probe")
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
en_rx_am.c was deleted in 'net-next' but had a bug fixed in it in
'net'.
The esp{4,6}_offload.c conflicts were overlapping changes.
The 'out' label is removed so we just return ERR_PTR(-EINVAL)
directly.
Signed-off-by: David S. Miller <davem@davemloft.net>
The trailing semicolon is an empty statement that does no operation.
Removing it since it doesn't do anything.
Signed-off-by: Luis de Bethencourt <luisbg@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Today 4 architectures set ARCH_SUPPORTS_MEMORY_FAILURE (arm64, parisc,
powerpc, and x86), while 4 other architectures set __ARCH_SI_TRAPNO
(alpha, metag, sparc, and tile). These two sets of architectures do
not interesect so remove the trapno paramater to remove confusion.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
With all the support code in place we can now link in the ipsec
offload operations and set the ESP feature flag for the XFRM
subsystem to see.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
If the skb has a security association referenced in the skb, then
set up the Tx descriptor with the ipsec offload bits. While we're
here, we fix an oddly named field in the context descriptor struct.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
If the chip sees and decrypts an ipsec offload, set up the skb
sp pointer with the ralated SA info. Since the chip is rude
enough to keep to itself the table index it used for the
decryption, we have to do our own table lookup, using the
hash for speed.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
On a chip reset most of the table contents are lost, so must be
restored. This scans the driver's ipsec tables and restores both
the filled and empty table slots to their pre-reset values.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add the functions for setting up and removing offloaded SAs (Security
Associations) with the x540 hardware. We set up the callback structure
but we don't yet set the hardware feature bit to be sure the XFRM service
won't actually try to use us for an offload yet.
The software tables are made up to mimic the hardware tables to make it
easier to track what's in the hardware, and the SA table index is used
for the XFRM offload handle. However, there is a hashing field in the
Rx SA tracking that will be used to facilitate faster table searches in
the Rx fast path.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add in the code for running and stopping the hardware ipsec
encryption/decryption engine. It is good to keep the engine
off when not in use in order to save on the power draw.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Pull networking fixes from David Miller:
1) Fix divide by zero in mlx5, from Talut Batheesh.
2) Guard against invalid GSO packets coming from untrusted guests and
arriving in qdisc_pkt_len_init(), from Eric Dumazet.
3) Similarly add such protection to the various protocol GSO handlers.
From Willem de Bruijn.
4) Fix regression added to IGMP source address checking for IGMPv3
reports, from Felix Feitkau.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
tls: Correct length of scatterlist in tls_sw_sendpage
be2net: restore properly promisc mode after queues reconfiguration
net: igmp: fix source address check for IGMPv3 reports
gso: validate gso_type in GSO handlers
net: qdisc_pkt_len_init() should be more robust
ibmvnic: Allocate and request vpd in init_resources
ibmvnic: Revert to previous mtu when unsupported value requested
ibmvnic: Modify buffer size and number of queues on failover
rds: tcp: compute m_ack_seq as offset from ->write_seq
usbnet: silence an unnecessary warning
cxgb4: fix endianness for vlan value in cxgb4_tc_flower
cxgb4: set filter type to 1 for ETH_P_IPV6
net/mlx5e: Fix fixpoint divide exception in mlx5e_am_stats_compare
Clean up the ipsec/macsec descriptor bit definitions to match the rest
of the defines and file organization. Also recognise the bit-definition
overlap in the error mask macro.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Although I'm not sure this parameter is useful for regular SRP users,
setting this parameter to 1 has shown to be invaluable for testing the
block layer core, SCSI core and device mapper queue running mechanisms.
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Since the SRP_LOGIN_REQ defined in the SRP standard is larger than
what fits in the RDMA/CM login request private data, introduce a new
login request format for the RDMA/CM.
Note: since srp_daemon and ibsrpdm rely on the subnet manager and
since there is no equivalent of the IB subnet manager in non-IB
networks, login has to be performed manually for non-IB networks.
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Pointers txr and rxr are being initialized and a few statements later
are being assigned new values without the original values ever being
read. The initialized values are therefore redundant and can be
removed.
Cleans up clang warnings:
drivers/net/ethernet/broadcom/bnx2.c:5821:28: warning: Value stored to
'txr' during its initialization is never read
drivers/net/ethernet/broadcom/bnx2.c:5822:28: warning: Value stored to
'rxr' during its initialization is never read
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We forgot to update the kernel doc header above sfp_register_upstream()
Fixes: c19bb00070 ("sfp: convert to fwnode")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes the following sparse warning:
drivers/net/ethernet/aquantia/atlantic/aq_pci_func.c:50:34: warning:
symbol 'hw_atl_boards' was not declared. Should it be static?
Fixes: 4948293ff9 ("net: aquantia: Introduce new AQC devices and capabilities")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix to return error code -ENOMEM from the aq_ndev_alloc() error
handling case instead of 0, as done elsewhere in this function.
Fixes: 23ee07ad3c ("net: aquantia: Cleanup pci functions module")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is possible to have CONFIG_OF enabled on x86 builds, where we have no
firmware provided max17042_platform_data. The CONFIG_OF implementation of
max17042_get_pdata would return NULL in this case, causing the probe to
fail.
Instead always fallback to the default platform-data, as used on x86 sofar,
when there is no firmware provided pdata, independent of CONFIG_OF.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
Fix to return error code -EINVAL instead of 0 when num_vfs above
limit_vfs, as done elsewhere in this function.
Fixes: 0dc7862191 ("nfp: handle SR-IOV already enabled when driver is probing")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
