With 64 VFs, we can easily overwhelm the AQ on the PF if we have too low
a limit on the number of AQ requests. This leads to ARQ overflow errors,
and occasionally VFs that fail to initialize.
Since we really only hit this condition on initial VF driver load, the
requests that we process are lightweight, so this extra work doesn't
cause problems for the PF driver.
Change-ID: I620221520d8af987df6ace9ba938ffaf22107681
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>
On some devices, in some systems, in some configurations, the VFs would
fail to initialize the first time you loaded the driver.
To correct this, increase the delay time for the init task slightly, and
wait longer before giving up.
If we enable VFs and load the VF driver in the same kernel as the PF
driver, we can totally overwhelm the PF driver with AQ requests because
all of the instances try to initialize at the same time.
To help alleviate this, stagger the initial scheduling of the init task
using the PCIe function as a multiplier. We mask off the function to
only three bits so no instance has to wait too long.
With these two changes, initializing 128 VFs on a single device goes
from four minutes to just a few seconds.
Change-ID: If3d8720c1c4e838ab36d8781d9ec295a62380936
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>
1000Base_T_Optical got added to the function that figures out what
is supported when link is down but not when link is up. Add it in there
too so that we display the correct information.
Change-ID: I85ebcdfa7c02d898c44c673b1500552a53c8042e
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The vlan_features field was correctly being set to the same value as the
netdev features field. However, this was being done before the features
were actually being set up, leaving the vlan_features empty.
Also, after a reset, vlan_features will be incorrectly assigned the
previous netdev feature flags, which can contain VLAN feature bits. This
makes the VLAN code angry and will cause a stack dump.
To fix these issues, set up the netdev features first, then mask out the
VLAN feature bits when assigning vlan_features.
Change-ID: Ib0548869dc83cf6a841cb8697dd94c12359ba4d2
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>
When the number of invalid messages from a VF is exceeded, the VF
will be disabled, due to the invalid messages. This happens if
other VF drivers (like DPDK) send a message through the driver's
mailbox (aka virtchannel) interface, but the message is not
supported by the i40e pf driver, such as CONFIG_PROMISCUOUS_MODE.
This patch changes the num_invalid_msgs in struct i40e_vf to record
the continuous invalid msgs, and it will be reset when a valid msg
is received.
Change-ID: Iaec42fd3dcdd281476b3518be23261dd46fc3718
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The XL710 hardware has a different interrupt moderation design
that can support a limit of total interrupts per second per
vector, in addition to the "number of interrupts per second"
controls already established in the driver. This combination
of hardware features allows us to set very low default latency
settings but minimize the total CPU utilization by not
making too many interrupts, should the user desire.
The current driver implementation is still enabling the dynamic
moderation in the driver, and only using the rx/tx-usecs
limit in ethtool to limit the interrupt rate per second, by default.
The new code implemented in this patch
2) adds init/use of the new "Interrupt Limit" register
3) adds ethtool knob to control/report the limits above
Usage is ethtool -C ethx rx-usecs-high <value> Where <value> is number
of microseconds to create a rate of 1/N interrupts per second,
regardless of rx-usecs or tx-usecs values. Since there is a credit based
scheme in the hardware, the rx-usecs and tx-usecs can be configured for
very low latency for short bursts, but once the credit runs out the
refill rate on the credits is limited by rx-usecs-high.
Change-ID: I3a1075d3296123b0f4f50623c779b027af5b188d
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Adds support for setting a new bit in the Set Local LLDP MIB AQ command
Type field. When set to 1, the bit indicates to FW that Apps should be
treated as non-willing. When 0, FW behaves as before.
Change-ID: I0d2101c1606c59c7188d3e6a0c7810e0f205233a
Signed-off-by: Greg Bowers <gregory.j.bowers@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add an ethtool priv flag to enable and disable printing
the VEB statistics.
Change-ID: I7654054a3a73b08aa8310d94ee8fce6219107dd8
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Two defines that are not used are causing customer confusion - remove
them.
Change-ID: Icef0325aca8e0f4fcdfc519e026bdd375e791200
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Allow the nvmupdate application to decide when a read or write error
should be exposed to the user. Since the application needs to use
write probes to find the ReadOnly sections on a potentially unknown NVM
version in the HW and read probes to check the status of the last write,
some error messages are expected, but need not be shown to the users.
The driver doesn't know which are ignorable from real errors, so needs
to let the application make the decision.
Change-ID: I78fca8ab672bede11c10c820b83c26adfd536d03
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add const to functions that return strings that aren't going to be
modified. This addresses some reported compile complaints.
Change-ID: Ic56b1e814ab4d23a50480e7fdec652445f776ee8
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cut down on the number of startup log entries by putting a couple behind
debug flags and combining a couple others into a single line.
Change-ID: I708089f086308f84d43f8b6f0e8a634a02d058fb
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
As per Eric Dumazet's previous patches:
(see commit (24d2e4a507) - tg3: use napi_complete_done())
Quoting verbatim:
Using napi_complete_done() instead of napi_complete() allows
us to use /sys/class/net/ethX/gro_flush_timeout
GRO layer can aggregate more packets if the flush is delayed a bit,
without having to set too big coalescing parameters that impact
latencies.
</end quote>
Tested
configuration: low latency via ethtool -C ethx adaptive-rx off
rx-usecs 10 adaptive-tx off tx-usecs 15
workload: streaming rx using netperf TCP_MAERTS
igb:
MIGRATED TCP MAERTS TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.0.1 () port 0 AF_INET : demo
...
Interim result: 941.48 10^6bits/s over 1.000 seconds ending at 1440193171.589
Alignment Offset Bytes Bytes Recvs Bytes Sends
Local Remote Local Remote Xfered Per Per
Recv Send Recv Send Recv (avg) Send (avg)
8 8 0 0 1176930056 1475.36 797726 16384.00 71905
MIGRATED TCP MAERTS TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.0.1 () port 0 AF_INET : demo
...
Interim result: 941.49 10^6bits/s over 0.997 seconds ending at 1440193142.763
Alignment Offset Bytes Bytes Recvs Bytes Sends
Local Remote Local Remote Xfered Per Per
Recv Send Recv Send Recv (avg) Send (avg)
8 8 0 0 1175182320 50476.00 23282 16384.00 71816
i40e:
Hard to test because the traffic is incoming so fast (24Gb/s) that GRO
always receives 87kB, even at the highest interrupt rate.
Other drivers were only compile tested.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The code in i40e and i40evf is using an "IN_NETPOLL" flag that has never
added any value due to the fact that the Rx clean-up is handled in NAPI.
As such the flag was set, the queue was scheduled via NAPI, and then polled
from the netpoll controller and if any Rx packets were processed the were
processed in the wrong context.
In addition the flag itself just added an unneeded conditional to the
hot-path so it can safely be dropped and save us a few instructions.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The polling routine for i40e was rounding up the budget for Rx cleanup to
1. This is incorrect as the netpoll poll call is expecting no Rx to be
processed as the budget passed was 0.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The device tree property "msi-map" specifies how to create the PCI
requester id used in some MSI controllers. Add a new function
of_msi_map_rid() that finds the msi-map property and applies its
translation to a given requester id.
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Currently msi-parent is used by a few bindings to describe the
relationship between a PCI root complex and a single MSI controller, but
this property does not have a generic binding document.
Additionally, msi-parent is insufficient to describe more complex
relationships between MSI controllers and devices under a root complex,
where devices may be able to target multiple MSI controllers, or where
MSI controllers use (non-probeable) sideband information to distinguish
devices.
This patch adds a generic binding for mapping PCI devices to MSI
controllers. This document covers msi-parent, and a new msi-map property
(specific to PCI*) which may be used to map devices (identified by their
Requester ID) to sideband data for each MSI controller that they may
target.
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
When we create a generic MSI domain, that MSI_FLAG_USE_DEF_CHIP_OPS
is set, and that any of .mask or .unmask are NULL in the irq_chip
structure, we set them to pci_msi_[un]mask_irq.
This is a bad idea for at least two reasons:
- PCI_MSI might not be selected, kernel fails to build (yes, this is
legitimate, at least on arm64!)
- This may not be a PCI/MSI domain at all (platform MSI, for example)
Either way, this looks wrong. Move the overriding of mask/unmask to
the PCI counterpart, and panic is any of these two methods is not
set in the core code (they really should be present).
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Link: http://lkml.kernel.org/r/1444760085-27857-1-git-send-email-marc.zyngier@arm.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
New users of these files still start showing up in linux-next, so it's
better to have a migration strategy. All existing users as of 4.3-rc4
are converted to use linux/io-64-nonatomic-*.h, and after 4.4-rc1
we can change all the new ones that have come in since, and then
remove this file again.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reported-by: LKP project <lkp@linux.intel.com>
to cover the entire kernel range. This fixes a triple fault on
non-PAE kernels when booting on 32-bit EFI due to accessing an
unmapped GDT in efi_call_phys_prolog() - Paolo Bonzini
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=2uOg
-----END PGP SIGNATURE-----
Merge tag 'efi-urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/efi into x86/urgent
Pull EFI fix from Matt Fleming:
- Ensure that the identity mapping in initial_page_table is updated
to cover the entire kernel range. This fixes a triple fault on
non-PAE kernels when booting on 32-bit EFI due to accessing an
unmapped GDT in efi_call_phys_prolog(). (Paolo Bonzini)
Signed-off-by: Ingo Molnar <mingo@kernel.org>
On 32-bit systems, the initial_page_table is reused by
efi_call_phys_prolog as an identity map to call
SetVirtualAddressMap. efi_call_phys_prolog takes care of
converting the current CPU's GDT to a physical address too.
For PAE kernels the identity mapping is achieved by aliasing the
first PDPE for the kernel memory mapping into the first PDPE
of initial_page_table. This makes the EFI stub's trick "just work".
However, for non-PAE kernels there is no guarantee that the identity
mapping in the initial_page_table extends as far as the GDT; in this
case, accesses to the GDT will cause a page fault (which quickly becomes
a triple fault). Fix this by copying the kernel mappings from
swapper_pg_dir to initial_page_table twice, both at PAGE_OFFSET and at
identity mapping.
For some reason, this is only reproducible with QEMU's dynamic translation
mode, and not for example with KVM. However, even under KVM one can clearly
see that the page table is bogus:
$ qemu-system-i386 -pflash OVMF.fd -M q35 vmlinuz0 -s -S -daemonize
$ gdb
(gdb) target remote localhost:1234
(gdb) hb *0x02858f6f
Hardware assisted breakpoint 1 at 0x2858f6f
(gdb) c
Continuing.
Breakpoint 1, 0x02858f6f in ?? ()
(gdb) monitor info registers
...
GDT= 0724e000 000000ff
IDT= fffbb000 000007ff
CR0=0005003b CR2=ff896000 CR3=032b7000 CR4=00000690
...
The page directory is sane:
(gdb) x/4wx 0x32b7000
0x32b7000: 0x03398063 0x03399063 0x0339a063 0x0339b063
(gdb) x/4wx 0x3398000
0x3398000: 0x00000163 0x00001163 0x00002163 0x00003163
(gdb) x/4wx 0x3399000
0x3399000: 0x00400003 0x00401003 0x00402003 0x00403003
but our particular page directory entry is empty:
(gdb) x/1wx 0x32b7000 + (0x724e000 >> 22) * 4
0x32b7070: 0x00000000
[ It appears that you can skate past this issue if you don't receive
any interrupts while the bogus GDT pointer is loaded, or if you avoid
reloading the segment registers in general.
Andy Lutomirski provides some additional insight:
"AFAICT it's entirely permissible for the GDTR and/or LDT
descriptor to point to unmapped memory. Any attempt to use them
(segment loads, interrupts, IRET, etc) will try to access that memory
as if the access came from CPL 0 and, if the access fails, will
generate a valid page fault with CR2 pointing into the GDT or
LDT."
Up until commit 23a0d4e8fa ("efi: Disable interrupts around EFI
calls, not in the epilog/prolog calls") interrupts were disabled
around the prolog and epilog calls, and the functional GDT was
re-installed before interrupts were re-enabled.
Which explains why no one has hit this issue until now. ]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reported-by: Laszlo Ersek <lersek@redhat.com>
Cc: <stable@vger.kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
[ Updated changelog. ]
Add helper function to handle the submission of fenced control requests.
Make sure we initialize the fence while holding the virtqueue lock, so
requests can't be reordered.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add virtio_gpu_queue_ctrl_buffer_locked function, which does the same as
virtio_gpu_queue_ctrl_buffer but does not take the virtqueue lock. The
caller must hold the lock instead.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Allow for arch-specific interrupt types to be set. For that, add
kvm_arch_set_irq() which takes interrupt type-specific action if it
recognizes the interrupt type given, and -EWOULDBLOCK otherwise.
The default implementation always returns -EWOULDBLOCK.
Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Vitaly Kuznetsov <vkuznets@redhat.com>
CC: "K. Y. Srinivasan" <kys@microsoft.com>
CC: Gleb Natapov <gleb@kernel.org>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Factor out kvm_notify_acked_gsi() helper to iterate over EOI listeners
and notify those matching the given gsi.
It will be reused in the upcoming Hyper-V SynIC implementation.
Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Vitaly Kuznetsov <vkuznets@redhat.com>
CC: "K. Y. Srinivasan" <kys@microsoft.com>
CC: Gleb Natapov <gleb@kernel.org>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The loop(for) inside irqfd_update() is unnecessary
because any other value for irq_entry.type will just trigger
schedule_work(&irqfd->inject) in irqfd_wakeup.
Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Vitaly Kuznetsov <vkuznets@redhat.com>
CC: "K. Y. Srinivasan" <kys@microsoft.com>
CC: Gleb Natapov <gleb@kernel.org>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As reported at https://bugs.launchpad.net/qemu/+bug/1494350,
it is possible to have vcpu->arch.st.last_steal initialized
from a thread other than vcpu thread, say the iothread, via
KVM_SET_MSRS.
Which can cause an overflow later (when subtracting from vcpu threads
sched_info.run_delay).
To avoid that, move steal time accumulation to vcpu entry time,
before copying steal time data to guest.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Reviewed-by: David Matlack <dmatlack@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Calling kvm_vcpu_gfn_to_memslot() twice in mapping_level() should be
avoided since getting a slot by binary search may not be negligible,
especially for virtual machines with many memory slots.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that it has only one caller, and its name is not so helpful for
readers, remove it. The new memslot_valid_for_gpte() function
makes it possible to share the common code between
gfn_to_memslot_dirty_bitmap() and mapping_level().
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is necessary to eliminate an extra memory slot search later.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As a bonus, an extra memory slot search can be eliminated when
is_self_change_mapping is true.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This will be passed to a function later.
Signed-off-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently we always write the next_rip of the shadow vmcb to
the guests vmcb when we emulate a vmexit. This could confuse
the guest when its cpuid indicated no support for the
next_rip feature.
Fix this by only propagating next_rip if the guest actually
supports it.
Cc: Bandan Das <bsd@redhat.com>
Cc: Dirk Mueller <dmueller@suse.com>
Tested-By: Dirk Mueller <dmueller@suse.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Expose VPID capability to L1. For nested guests, we don't do anything
specific for single context invalidation. Hence, only advertise support
for global context invalidation. The major benefit of nested VPID comes
from having separate vpids when switching between L1 and L2, and also
when L2's vCPUs not sched in/out on L1.
Reviewed-by: Wincy Van <fanwenyi0529@gmail.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
VPID is used to tag address space and avoid a TLB flush. Currently L0 use
the same VPID to run L1 and all its guests. KVM flushes VPID when switching
between L1 and L2.
This patch advertises VPID to the L1 hypervisor, then address space of L1
and L2 can be separately treated and avoid TLB flush when swithing between
L1 and L2. For each nested vmentry, if vpid12 is changed, reuse shadow vpid
w/ an invvpid.
Performance:
run lmbench on L2 w/ 3.5 kernel.
Context switching - times in microseconds - smaller is better
-------------------------------------------------------------------------
Host OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw
--------- ------------- ------ ------ ------ ------ ------ ------- -------
kernel Linux 3.5.0-1 1.2200 1.3700 1.4500 4.7800 2.3300 5.60000 2.88000 nested VPID
kernel Linux 3.5.0-1 1.2600 1.4300 1.5600 12.7 12.9 3.49000 7.46000 vanilla
Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Wincy Van <fanwenyi0529@gmail.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add the INVVPID instruction emulation.
Reviewed-by: Wincy Van <fanwenyi0529@gmail.com>
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Because eth_type_trans() consumes ethernet header worth of bytes, a call
to read TCI from end of packet using rhine_rx_vlan_tag() no longer works
as it's reading from an invalid offset.
Tested to be working on PCEngines Alix board.
Fixes: 810f19bcb8 ("via-rhine: add consistent memory barrier in vlan receive code.")
Signed-off-by: Andrej Ota <andrej@ota.si>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, is only called from __prog_put_rcu in the bpf_prog_release
path. Need this to call this from bpf_prog_put also to get correct
accounting.
Fixes: aaac3ba95e ("bpf: charge user for creation of BPF maps and programs")
Signed-off-by: Tom Herbert <tom@herbertland.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp/dccp: make our listener code more robust
This patch series addresses request sockets leaks and listener dismantle
phase. This survives a stress test with listeners being added/removed
quite randomly.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
