By giving each register its own liveness chain, we elide the skip_callee()
 logic.  Instead, each register's parent is the state it inherits from;
 both check_func_call() and prepare_func_exit() automatically connect
 reg states to the correct chain since when they copy the reg state across
 (r1-r5 into the callee as args, and r0 out as the return value) they also
 copy the parent pointer.
Signed-off-by: NEdward Cree <ecree@solarflare.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

Björn Töpel says:

====================
This patch set introduces zero-copy AF_XDP support for Intel's i40e
driver. In the first preparatory patch we also add support for
XDP_REDIRECT for zero-copy allocated frames so that XDP programs can
redirect them. This was a ToDo from the first AF_XDP zero-copy patch
set from early June. Special thanks to Alex Duyck and Jesper Dangaard
Brouer for reviewing earlier versions of this patch set.

The i40e zero-copy code is located in its own file i40e_xsk.[ch]. Note
that in the interest of time, to get an AF_XDP zero-copy implementation
out there for people to try, some code paths have been copied from the
XDP path to the zero-copy path. It is out goal to merge the two paths
in later patch sets.

In contrast to the implementation from beginning of June, this patch
set does not require any extra HW queues for AF_XDP zero-copy
TX. Instead, the XDP TX HW queue is used for both XDP_REDIRECT and
AF_XDP zero-copy TX.

Jeff, given that most of changes are in i40e, it is up to you how you
would like to route these patches. The set is tagged bpf-next, but
if taking it via the Intel driver tree is easier, let us know.

We have run some benchmarks on a dual socket system with two Broadwell
E5 2660 @ 2.0 GHz with hyperthreading turned off. Each socket has 14
cores which gives a total of 28, but only two cores are used in these
experiments. One for TR/RX and one for the user space application. The
memory is DDR4 @ 2133 MT/s (1067 MHz) and the size of each DIMM is
8192MB and with 8 of those DIMMs in the system we have 64 GB of total
memory. The compiler used is gcc (Ubuntu 7.3.0-16ubuntu3) 7.3.0. The
NIC is Intel I40E 40Gbit/s using the i40e driver.

Below are the results in Mpps of the I40E NIC benchmark runs for 64
and 1500 byte packets, generated by a commercial packet generator HW
outputing packets at full 40 Gbit/s line rate. The results are with
retpoline and all other spectre and meltdown fixes, so these results
are not comparable to the ones from the zero-copy patch set in June.

AF_XDP performance 64 byte packets.
Benchmark   XDP_SKB    XDP_DRV    XDP_DRV with zerocopy
rxdrop       2.6        8.2         15.0
txpush       2.2        -           21.9
l2fwd        1.7        2.3         11.3

AF_XDP performance 1500 byte packets:
Benchmark   XDP_SKB   XDP_DRV     XDP_DRV with zerocopy
rxdrop       2.0        3.3         3.3
l2fwd        1.3        1.7         3.1

XDP performance on our system as a base line:

64 byte packets:
XDP stats       CPU     pps         issue-pps
XDP-RX CPU      16      18.4M  0

1500 byte packets:
XDP stats       CPU     pps         issue-pps
XDP-RX CPU      16      3.3M    0

The structure of the patch set is as follows:

Patch 1: Add support for XDP_REDIRECT of zero-copy allocated frames
Patches 2-4: Preparatory patches to common xsk and net code
Patches 5-7: Preparatory patches to i40e driver code for RX
Patch 8: i40e zero-copy support for RX
Patch 9: Preparatory patch to i40e driver code for TX
Patch 10: i40e zero-copy support for TX
Patch 11: Add flags to sample application to force zero-copy/copy mode
====================
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

The -c/--copy -z/--zero-copy flags enforces either copy or zero-copy
mode.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

This patch adds zero-copy Tx support for AF_XDP sockets. It implements
the ndo_xsk_async_xmit netdev ndo and performs all the Tx logic from a
NAPI context. This means pulling egress packets from the Tx ring,
placing the frames on the NIC HW descriptor ring and completing sent
frames back to the application via the completion ring.

The regular XDP Tx ring is used for AF_XDP as well. This rationale for
this is as follows: XDP_REDIRECT guarantees mutual exclusion between
different NAPI contexts based on CPU id. In other words, a netdev can
XDP_REDIRECT to another netdev with a different NAPI context, since
the operation is bound to a specific core and each core has its own
hardware ring.

As the AF_XDP Tx action is running in the same NAPI context and using
the same ring, it will also be protected from XDP_REDIRECT actions
with the exact same mechanism.

As with AF_XDP Rx, all AF_XDP Tx specific functions are added to
i40e_xsk.c.
Signed-off-by: NMagnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

This patch prepares for the upcoming zero-copy Tx functionality, by
moving common functions and refactor chunks of code into re-usable
functions, used both by the regular path and zero-copy path.
Signed-off-by: NMagnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

This patch adds zero-copy Rx support for AF_XDP sockets. Instead of
allocating buffers of type MEM_TYPE_PAGE_SHARED, the Rx frames are
allocated as MEM_TYPE_ZERO_COPY when AF_XDP is enabled for a certain
queue.

All AF_XDP specific functions are added to a new file, i40e_xsk.c.

Note that when AF_XDP zero-copy is enabled, the XDP action XDP_PASS
will allocate a new buffer and copy the zero-copy frame prior passing
it to the kernel stack.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

This patch prepares for the upcoming zero-copy Rx functionality, by
moving/changing linkage of common functions, used both by the regular
path and zero-copy path.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

In this commit, the Rx path is refactored some, as a step torwards the
introduction AF_XDP Rx zero-copy.

The page re-use counter is moved into the i40e_reuse_rx_page, instead
of bumping the counter in many places. The Rx buffer page clearing is
moved for better readability. Lastely, functions to update statistics
and bump the XDP Tx ring are introduced.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

Add functions for queue pair enable/disable. Instead of resetting the
whole device, only the affected queue pair is disabled or enabled.

This plumbing is used in a later commit, when zero-copy AF_XDP support
is introduced.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

The function napi_if_scheduled_mark_missed is used to check if the
NAPI context is scheduled, if so set NAPIF_STATE_MISSED and return
true. Used by the AF_XDP zero-copy i40e Tx code implementation in
order to make sure that irq affinity is honored by the napi context.
Signed-off-by: NMagnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

Move the xdp_umem_get_{data,dma} functions to include/net/xdp_sock.h,
so that the upcoming zero-copy implementation in the Ethernet drivers
can utilize them.

Also, supply some dummy function implementations for
CONFIG_XDP_SOCKETS=n configs.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

Export __xdp_rxq_info_unreg_mem_model as xdp_rxq_info_unreg_mem_model,
so it can be used from netdev drivers. Also, add additional checks for
the memory type.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

This commit adds proper MEM_TYPE_ZERO_COPY support for
convert_to_xdp_frame. Converting a MEM_TYPE_ZERO_COPY xdp_buff to an
xdp_frame is done by transforming the MEM_TYPE_ZERO_COPY buffer into a
MEM_TYPE_PAGE_ORDER0 frame. This is costly, and in the future it might
make sense to implement a more sophisticated thread-safe alloc/free
scheme for MEM_TYPE_ZERO_COPY, so that no allocation and copy is
required in the fast-path.
Signed-off-by: NBjörn Töpel <bjorn.topel@intel.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>

If the user supplies a --cgroup value in the arguments when running
the test_suite go ahaead and run the self tests there. I use this
to test with multiple cgroup users.
Signed-off-by: NJohn Fastabend <john.fastabend@gmail.com>
Signed-off-by: NDaniel Borkmann <daniel@iogearbox.net>

Currently, we do a shutdown(sk, SHUT_RDWR) on both peer sockets and
a shutdown on the sender as well. However, this is incorrect and can
occasionally cause issues if you happen to have bad timing. First
peer1 or peer2 may still be in use depending on the test and timing.
Second we really should only be closing the read side and/or write
side depending on if the test is receiving or sending.

But, really none of this is needed just remove the shutdown calls.
Signed-off-by: NJohn Fastabend <john.fastabend@gmail.com>
Signed-off-by: NDaniel Borkmann <daniel@iogearbox.net>

Remove duplicated include.
Signed-off-by: NYueHaibing <yuehaibing@huawei.com>
Signed-off-by: NDaniel Borkmann <daniel@iogearbox.net>

Jakub Kicinski says:

====================
nfp: add NFP5000 support

This series broadly speaking adds support for NFP5000 and
related products.

First we add support for loading FW from flash.  We need to allow
for the management processor to provide extended log messages when
FW is loaded.  This is needed when FW selection policy is to compare
the FW on the disk and in the flash, and load the newer.  User should
be told what FW was selected.

We use this opportunity to add extended errors for normal FW loading
as well.

Next we add support for requesting HW information from the management
processor.  Up until now the driver read the HWinfo as it appears in
card memory, but there can be cases when management processor has
additional information or generates the entries dynamically so
occasionally we will have to consult it.  We use this to look up MAC
addresses for PCIe netdevs.

Next the actual patch with NFP5000 support and a small dose of
refactoring of PCIe init.

The remaining patches add support for reading RTsymbol types we
didn't need before.  Ones explicitly placed in external memory unit's
cache and absolute ones.

This part begins with a patch moving the logic which figures out
the correct bit offsets to device probe, to avoid redoing the
calculation for each access.  Second patch adds error messages
for easier troubleshooting.  Next patch adds helpers which will
take care of address conversions to reach into EMU cache.
Subsequently users are migrated from the raw CPP API to the new RTsym
helpers.  Finally we add support for reading absolute symbols.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make the RTsym users access the size via the helper, which
takes care of special handling of absolute symbols.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add support in nfpcore for reading the absolute RTsyms.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Convert all users of RTsym to the new set of helpers which
handle all targets correctly.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make nfp_rtsym_{read,write}_le() and nfp_rtsym_map() use the new
target resolution helpers to allow accessing in-cache symbols.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Align nfp_cpp_map_area() with other CPP-level APIs and pass
encoded cpp_id/dest rather than target, action, domain tuple.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RTsyms may have special encodings for more complex symbol types.
For example symbols which are placed in external memory unit's
cache directly, constants or local memory.  Add set of helpers
which will check for those special encodings and handle them
correctly.

For now only add direct cache accesses, we don't have a need to
access the other ones in foreseeable future.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add error prints to CPP target encoding/decoding logic, otherwise
it's quite hard to pin point the reasons why read or write
operations fail.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We will soon need the MU locality field offset much more
often than just for decoding MIP address.  Save it in nfp_cpp
for quick access.  Note that we can already reuse the target
config from nfp_cpp, no need to do the XPB read.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NFrancois H. Theron <francois.theron@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use a switch statement instead of ifs for code dependent
on chip version.  While at it make sure we fail for unknown
chip revisions.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add NFP5000 to supported chips, the chip is backward compatible
with NFP4000 and NFP6000, so core PCIe code needs to handle it
the same way as 4k and 6k.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In multi-host scenarios Management FW may allocate MAC addresses
at runtime, we have to use the indirect lookup to find them.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NDirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Management FW can adjust some of the information in the HWinfo table
at runtime.  In some cases reading the table directly will not yield
correct results.  Add a NSP command for looking up information.
Up until now we weren't making use of any of the values which may
get adjusted.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NDirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To enable easier FW distribution NFP can now automatically
select between FW stored on the flash and loaded from the
kernel.

If FW loading policy is set to auto it will compare the
versions of FW from the host and from the flash and load
the newer one.  If FW type doesn't match (e.g. one advanced
application vs another) the FW from the host takes precedence,
unless one of them is the basic NIC firmware, in which case
the non-basic-NIC FW is selected.

This automatic selection mechanism requires we inform user
what the verdict was.  Print a message to the logs explaining
the decision and the reason.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NDirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Flash may contain a default NFP application FW.  This application
can either be put there by the user (with ethtool -f) or shipped
with the card.  If file system FW is not found, attempt to load
this flash stored app FW.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NDirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is already a fair number of arguments to nfp_nsp_command()
family of functions.  Encapsulate them into structures to make
adding new ones easier.  No functional changes.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NDirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2018-08-28

This series contains new features and implementation updates for the
ice driver.

Anirudh reworks the current flex programming logic to add support for
a second flex descriptor profile.  Updated the transmit scheduler
code to handle changes to the spec, specifically the firmware expects
a 4KB buffer at all times so fix the default scheduler topology buffer
size.  Also the maximum children per node per layer is replaced by
maximum sibling group size.  Adds a check to ensure a reset is not in
progress before exercising a control queue operation.  Refactored the
switch rule management functions and structures to simply the logic and
to add a common function to search for a rule entry and add a new rule
entry.  Refactored the VSI allocation, deletion and rebuild flow so that
on reset we can restore all the filters that were previously added.  Did
some spring cleaning of define names and macros.

Dan updates the admin queue command for requesting resource ownership
to the latest specification by adding new enum's and change the locks.

Zhenning optimizes the driver by using the existing buffer in a
structure directly versus a local array.

Chinh implements handlers for ethtool for get and set link settings.

Sudheer implements transmit hang/timeout detection and malicious driver
detection in the driver.

Md Fahad Iqbal implements the get and set bridge mode operations.

Hieu adds the ability for firmware logging during initialization.

Brett updates the driver to only enable VSI transmit and receive pruning
when VLAN 0 is active, and when VLAN 0 is removed/not active, pruning is
disabled.

Akeem adds a flag to use for stopping the service task.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
10GbE Intel Wired LAN Driver Updates 2018-08-28

This series contains updates to ixgbe and ixgbevf only.

Sebastian adds support for firmware NVM recovery mode, which logs a
message when errors are detected and un-registers the device.  Also
fixed RSS type recognition with VF to VF communication.

Shannon Nelson implements IPsec hardware offload for VF devices in
Intel's 10GbE x540 family of Ethernet devices.

The IPsec HW offload feature has been in the x540/Niantic family of
network devices since their release in 2009, but there was no Linux
kernel support for the offload until 2017.  After the XFRM code added
support for the offload last year, the HW offload was added to the ixgbe
PF driver.

Since the related x540 VF device uses same setup as the PF for implementing
the offload, adding the feature to the ixgbevf seemed like a good idea.
In this case, the PF owns the device registers, so the VF simply packages
up the request information into a VF<->PF message and the PF does the
device configuration.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When failing the request because we can't support that offload,
reporting EOPNOTSUPP makes much more sense than ENXIO.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

There seems to be a problem in the x540's internal switch wherein if SR-IOV
mode is enabled and an offloaded IPsec packet is sent to a local VF,
the packet is silently dropped.  This might never be a problem as it is
somewhat a corner case, but if someone happens to be using IPsec offload
from the PF to a VF that just happens to get migrated to the local box,
communication will mysteriously fail.

Not good.

A simple way to protect from this is to simply not allow any IPsec offloads
for outgoing packets when num_vfs != 0.  This doesn't help any offloads that
were created before SR-IOV was enabled, but we'll get to that later.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add the IPsec initialization into the driver startup and
add the Rx and Tx processing hooks.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add the IPsec offload support code.  This is based off of the similar
code in ixgbe, but instead of writing the SA registers, the VF asks
the PF to setup the offload by sending the offload information to the
PF via the standard mailbox.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Fix up the register definitions for using IPsec offloads and
add the new mailbox message IDs.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add an add and a delete message for IPsec offload requests from
the VF.  These call into the IPsec functions that can translate
the message buffer into a useful IPsec offload.

These new messages bump the mbox API version to 1.4.
Signed-off-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>