Move the received_rps counter value next to the other RPS-related members
in softnet_data. This closes two four-byte holes in the structure, making
room for another pointer in the first two cache lines without bumping the
xmit struct to its own line.
Acked-by: Song Liu <song@kernel.org>
Reviewed-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20221108140601.149971-2-toke@redhat.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

This change documents the process for running the BPF CI before
submitting a patch to the upstream mailing list, similar to what happens
if a patch is send to bpf@vger.kernel.org: it builds kernel and
selftests and runs the latter on different architecture (but it notably
does not cover stylistic checks such as cover letter verification).

Running BPF CI this way can help achieve better test coverage ahead of
patch submission than merely running locally (say, using
tools/testing/selftests/bpf/vmtest.sh), as additional architectures may
be covered as well.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20221114211501.2068684-1-deso@posteo.net

Instead of having to pass multiple arguments that describe the register,
pass the bpf_reg_state into the btf_struct_access callback. Currently,
all call sites simply reuse the btf and btf_id of the reg they want to
check the access of. The only exception to this pattern is the callsite
in check_ptr_to_map_access, hence for that case create a dummy reg to
simulate PTR_TO_BTF_ID access.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-8-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

Currently, verifier uses MEM_ALLOC type tag to specially tag memory
returned from bpf_ringbuf_reserve helper. However, this is currently
only used for this purpose and there is an implicit assumption that it
only refers to ringbuf memory (e.g. the check for ARG_PTR_TO_ALLOC_MEM
in check_func_arg_reg_off).

Hence, rename MEM_ALLOC to MEM_RINGBUF to indicate this special
relationship and instead open the use of MEM_ALLOC for more generic
allocations made for user types.

Also, since ARG_PTR_TO_ALLOC_MEM_OR_NULL is unused, simply drop it.

Finally, update selftests using 'alloc_' verifier string to 'ringbuf_'.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-7-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

Currently, the verifier has two return types, RET_PTR_TO_ALLOC_MEM, and
RET_PTR_TO_ALLOC_MEM_OR_NULL, however the former is confusingly named to
imply that it carries MEM_ALLOC, while only the latter does. This causes
confusion during code review leading to conclusions like that the return
value of RET_PTR_TO_DYNPTR_MEM_OR_NULL (which is RET_PTR_TO_ALLOC_MEM |
PTR_MAYBE_NULL) may be consumable by bpf_ringbuf_{submit,commit}.

Rename it to make it clear MEM_ALLOC needs to be tacked on top of
RET_PTR_TO_MEM.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-6-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

Add the support on the map side to parse, recognize, verify, and build
metadata table for a new special field of the type struct bpf_list_head.
To parameterize the bpf_list_head for a certain value type and the
list_node member it will accept in that value type, we use BTF
declaration tags.

The definition of bpf_list_head in a map value will be done as follows:

struct foo {
	struct bpf_list_node node;
	int data;
};

struct map_value {
	struct bpf_list_head head __contains(foo, node);
};

Then, the bpf_list_head only allows adding to the list 'head' using the
bpf_list_node 'node' for the type struct foo.

The 'contains' annotation is a BTF declaration tag composed of four
parts, "contains:name:node" where the name is then used to look up the
type in the map BTF, with its kind hardcoded to BTF_KIND_STRUCT during
the lookup. The node defines name of the member in this type that has
the type struct bpf_list_node, which is actually used for linking into
the linked list. For now, 'kind' part is hardcoded as struct.

This allows building intrusive linked lists in BPF, using container_of
to obtain pointer to entry, while being completely type safe from the
perspective of the verifier. The verifier knows exactly the type of the
nodes, and knows that list helpers return that type at some fixed offset
where the bpf_list_node member used for this list exists. The verifier
also uses this information to disallow adding types that are not
accepted by a certain list.

For now, no elements can be added to such lists. Support for that is
coming in future patches, hence draining and freeing items is done with
a TODO that will be resolved in a future patch.

Note that the bpf_list_head_free function moves the list out to a local
variable under the lock and releases it, doing the actual draining of
the list items outside the lock. While this helps with not holding the
lock for too long pessimizing other concurrent list operations, it is
also necessary for deadlock prevention: unless every function called in
the critical section would be notrace, a fentry/fexit program could
attach and call bpf_map_update_elem again on the map, leading to the
same lock being acquired if the key matches and lead to a deadlock.
While this requires some special effort on part of the BPF programmer to
trigger and is highly unlikely to occur in practice, it is always better
if we can avoid such a condition.

While notrace would prevent this, doing the draining outside the lock
has advantages of its own, hence it is used to also fix the deadlock
related problem.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-5-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

The current offset needs to also skip over the already copied region in
addition to the size of the next field. This case manifests where there
are gaps between adjacent special fields.

It was observed that for a map value with size 48, having fields at:
off:  0, 16, 32
size: 4, 16, 16

The current code does:

memcpy(dst + 0, src + 0, 0)
memcpy(dst + 4, src + 4, 12)
memcpy(dst + 20, src + 20, 12)
memcpy(dst + 36, src + 36, 12)

With the fix, it is done correctly as:

memcpy(dst + 0, src + 0, 0)
memcpy(dst + 4, src + 4, 12)
memcpy(dst + 32, src + 32, 0)
memcpy(dst + 48, src + 48, 0)

Fixes: 4d7d7f69 ("bpf: Adapt copy_map_value for multiple offset case")
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-4-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

In f71b2f64 ("bpf: Refactor map->off_arr handling"), map->off_arr
was refactored to be btf_field_offs. The number of field offsets is
equal to maximum possible fields limited by BTF_FIELDS_MAX. Hence, reuse
BTF_FIELDS_MAX as spin_lock and timer no longer are to be handled
specially for offset sorting, fix the comment, and remove incorrect
WARN_ON as its rec->cnt can never exceed this value. The reason to keep
separate constant was the it was always more 2 more than total kptrs.
This is no longer the case.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-3-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

We don't want to commit to a specific name for these. Simply call them
allocated objects coming from bpf_obj_new, which is completely clear in
itself.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221114191547.1694267-2-memxor@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

Kang Minchul says:

====================
This patch series contains various checkpatch fixes
in btf.c, libbpf.c, ringbuf.c.

I know these are trivial but some issues are hard to ignore
and I think these checkpatch issues are accumulating.

v1 -> v2: changed cover letter message.
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>

Fixed some checkpatch issues in ringbuf.c
Signed-off-by: Kang Minchul <tegongkang@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20221113190648.38556-4-tegongkang@gmail.com

Fixed following checkpatch issues:

WARNING: Block comments use a trailing */ on a separate line
+        * other BPF program's BTF object */

WARNING: Possible repeated word: 'be'
+        * name. This is important to be be able to find corresponding BTF

ERROR: switch and case should be at the same indent
+       switch (ext->kcfg.sz) {
+               case 1: *(__u8 *)ext_val = value; break;
+               case 2: *(__u16 *)ext_val = value; break;
+               case 4: *(__u32 *)ext_val = value; break;
+               case 8: *(__u64 *)ext_val = value; break;
+               default:

ERROR: trailing statements should be on next line
+               case 1: *(__u8 *)ext_val = value; break;

ERROR: trailing statements should be on next line
+               case 2: *(__u16 *)ext_val = value; break;

ERROR: trailing statements should be on next line
+               case 4: *(__u32 *)ext_val = value; break;

ERROR: trailing statements should be on next line
+               case 8: *(__u64 *)ext_val = value; break;

ERROR: code indent should use tabs where possible
+                }$

WARNING: please, no spaces at the start of a line
+                }$

WARNING: Block comments use a trailing */ on a separate line
+        * for faster search */

ERROR: code indent should use tabs where possible
+^I^I^I^I^I^I        &ext->kcfg.is_signed);$

WARNING: braces {} are not necessary for single statement blocks
+       if (err) {
+               return err;
+       }

ERROR: code indent should use tabs where possible
+^I^I^I^I        sizeof(*obj->btf_modules), obj->btf_module_cnt + 1);$
Signed-off-by: Kang Minchul <tegongkang@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20221113190648.38556-3-tegongkang@gmail.com

Fixed some checkpatch issues in btf.c
Signed-off-by: Kang Minchul <tegongkang@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20221113190648.38556-2-tegongkang@gmail.com

Fixup bpf_map_update_elem() declaration to use a single line.
Reported-by: Akira Yokosawa <akiyks@gmail.com>
Signed-off-by: Maryam Tahhan <mtahhan@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Akira Yokosawa <akiyks@gmail.com>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20221113103327.3287482-1-mtahhan@redhat.com

GCC 11.3.0 fails to compile btf_dump.c due to the following error,
which seems to originate in btf_dump_struct_data where the returned
value would be uninitialized if btf_vlen returns zero.

btf_dump.c: In function ‘btf_dump_dump_type_data’:
btf_dump.c:2363:12: error: ‘err’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
 2363 |         if (err < 0)
      |            ^

Fixes: 920d16af ("libbpf: BTF dumper support for typed data")
Signed-off-by: David Michael <fedora.dm0@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Stanislav Fomichev <sdf@google.com>
Acked-by: Alan Maguire <alan.maguire@oracle.com>
Link: https://lore.kernel.org/bpf/87zgcu60hq.fsf@gmail.com

Saeed Mahameed says:

====================
mlx5-updates-2022-11-12

Misc updates to mlx5 driver

1) Support enhanced CQE compression, on ConnectX6-Dx
   Reduce irq rate, cpu utilization and latency.

2) Connection tracking: Optimize the pre_ct table lookup for rules
   installed on chain 0.

3) implement ethtool get_link_ext_stats for PHY down events

4) Expose device vhca_id to debugfs

5) misc cleanups and trivial changes
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Added xdp and page pool statistics.
In order to make the implementation simple and compatible, the patch
uses the 32bit integer to record the XDP statistics.
Signed-off-by: Shenwei Wang <shenwei.wang@nxp.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Steen Hegelund says:

====================
net: Add support for sorted VCAP rules in Sparx5

This provides support for adding Sparx5 VCAP rules in sorted order, VCAP
rule counters and TC filter matching on ARP frames.

It builds on top of the initial IS2 VCAP support found in these series:

https://lore.kernel.org/all/20221020130904.1215072-1-steen.hegelund@microchip.com/
https://lore.kernel.org/all/20221109114116.3612477-1-steen.hegelund@microchip.com/

Functionality
=============

When a new VCAP rule is added the driver will now ensure that the rule is
inserted in sorted order, and when a rule is removed, the remaining rules
will be moved to keep the sorted order and remove any gaps in the VCAP
address space.

A VCAP rule is ordered using these 3 values:

 - Rule size: the count of VCAP addresses used by the rule.  The largest
   rule have highest priority

 - Rule User: The rules are ordered by the user enumeration

 - Priority: The priority provided in the flower filter.  The lowest value
   has the highest priority.

A VCAP instance may contain the counter as part of the VCAP cache area, and
this counter may be one or more bits in width.  This type of counter
automatically increments its value when the rule is hit.

Other VCAP instances have a dedicated counter area outside of the VCAP and
in this case the rule must contain the counter id to be able to locate the
counter value and cause the counter to be incremented.  In this case there
must also be a VCAP rule action that sets the counter id.

The Sparx5 IS2 VCAP uses a dedicated counter area with 32bit counters.

This series adds support for getting VCAP rule counters and provide these
via the TC statistic interface.

This only support packet counters, not byte counters.

Finally the series adds support for the ARP frame dissector and configures
the Sparx5 IS2 VCAP to generate the ARP keyset when ARP traffic is
received.

Delivery:
=========

This is current plan for delivering the full VCAP feature set of Sparx5:

- DebugFS support for inspecting rules
- TC protocol all support
- Sparx5 IS0 VCAP support
- TC policer and drop action support (depends on the Sparx5 QoS support
  upstreamed separately)
- Sparx5 ES0 VCAP support
- TC flower template support
- TC matchall filter support for mirroring and policing ports
- TC flower filter mirror action support
- Sparx5 ES2 VCAP support
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This tests the insert, move and deleting of rules and checks that the
unused VCAP addresses are initialized correctly.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This provides flower filter packet statistics (bytes are not supported) via
the dedicated IS2 counter feature.

All rules having the same TC cookie will contribute to the packet
statistics for the filter as they are considered to be part of the same TC
flower filter.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This adds API methods to set and get a rule counter.

A VCAP instance may contain the counter as part of the VCAP cache area, and
this counter may be one or more bits in width.  This type of counter
automatically increments it value when the rule is hit.

Other VCAP instances have a dedicated counter area outside of the VCAP and
in this case the rule must contain the counter id to be able to locate the
counter value.  In this case there must also be a rule action that updates
the counter using the rule id when the rule is hit.

The Sparx5 IS2 VCAP uses a dedicated counter area.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This adds a sorting criteria to rule insertion and deletion.

The criteria is (in the listed order):

- Rule size (largest size first)
- User (based on an enumerated user value)
- Priority (highest priority first, aka lowest value)

When a rule is deleted the other rules may need to be moved to fill the gap
to use the available VCAP address space in the best possible way.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This add support for Sparx5 for dissecting TC ARP flower filter keys and
sets up the Sparx5 IS2 VCAP to generate the ARP keyset for ARP frames.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This adds a new flow_rule_match_arp function that allows drivers
to be able to dissect ARP frames.
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The value of 'st->state' has been verified as "TCP_SEQ_STATE_LISTENING",
it's unnecessary to assign TCP_SEQ_STATE_LISTENING to it, so we can remove it.
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

Nick Child says:

====================
ibmvnic: Introduce affinity hint support

This is a patchset to do 3 things to improve ibmvnic performance:
    1. Assign affinity hints to ibmvnic queue irq's
    2. Update affinity hints on cpu hotplug events
    3. Introduce transmit packet steering (XPS)

NOTE: If irqbalance is running, you need to stop it from overriding
  our affinity hints. To do this you can do one of:
   - systemctl stop irqbalance
   - ban the ibmvnic module irqs
      - you must have the latest irqbalance v9.2, the banmod argument was broken before this
      - in /etc/sysconfig/irqbalance -> IRQBALANCE_ARGS="--banmod=ibmvnic"
      - systemctl restart irqbalance
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Transmit Packet Steering (XPS) maps cpu numbers to transmit
queues. By running the same connection on the same set of cpu's,
contention for the queue and cache miss rate can be minimized.
When assigning a cpu mask for a tranmit queues irq number, assign
the same cpu mask as the set of cpu's that XPS should use for that
queue.
Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Rick Lindsley <ricklind@linux.ibm.com>
Reviewed-by: Haren Myneni <haren@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When CPU's are added and removed, ibmvnic devices will reassign
hint values. Introduce a new cpu hotplug state CPUHP_IBMVNIC_DEAD
to signal to ibmvnic devices that the CPU has been removed and it
is time to reset affinity hint assignments. On the other hand,
when CPU's are being added, add a state instance to
CPUHP_AP_ONLINE_DYN which will trigger a reassignment of affinity
hints once the new CPU's are online. This implementation is based
on the virtio_net driver.
Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Rick Lindsley <ricklind@linux.ibm.com>
Reviewed-by: Haren Myneni <haren@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Assign affinity hints to ibmvnic device queue interrupts.
Affinity hints are assigned and removed during sub-crq init and
teardown, respectively. This update should improve latency if
utilized as interrupt lines and processing are more equally
distributed among CPU's. This implementation is based on the
virtio_net driver.
Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Rick Lindsley <ricklind@linux.ibm.com>
Reviewed-by: Haren Myneni <haren@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for 2 virtchnl msgs:
VIRTCHNL_OP_SET_RSS_HENA
VIRTCHNL_OP_GET_RSS_HENA_CAPS

The first one allows VFs to clear all previously programmed
RSS configuration and customize it. The second one returns
the RSS HENA bits allowed by the hardware.

Introduce ice_err_to_virt_err which converts kernel
specific errors to virtchnl errors.
Signed-off-by: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The error handling in tun_get_user is very scattered.
This patch unifies error handling, reduces duplication of code, and
makes the logic clearer.
Signed-off-by: Chuang Wang <nashuiliang@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implement ethtool_op get_link_ext_stats for PHY down events
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>

The pre_ct table realizes in hardware the act_ct cache logic, bypassing
the CT table if the ct state was already set by a previous ct lookup.
As such, the pre_ct table will always miss for chain 0 filters.

Optimize the pre_ct table lookup for rules installed on chain 0.
Signed-off-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

A single async context object is sufficient to wait for the completions
of many callbacks.  Switch to using one instance per a bulk of commands.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Waiting on a completion object for each callback before cleaning up their
async contexts is not necessary, as this is already implied in the
mlx5_cmd_cleanup_async_ctx() API.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Work field in struct mlx5e_async_ctx is not used. Remove it.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

The case where the packet is not offloaded and needs to be restored
to slow path and couldn't find expected tunnel information should not
dump a call trace to the user. there is a debug call.
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Up until now, return value of update_rx was ignored. Therefore, flow
continues even if it fails. Add error flow in case of update_rx fails in
mlx5e_open_locked, mlx5i_open and mlx5i_pkey_open.
Signed-off-by: Guy Truzman <gtruzman@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Params info print was meant to be printed on load.
With time, new calls to mlx5e_init_rq_type_params and
mlx5e_build_rq_params were added, mistakenly printing
the params once again.

Move the print to were it belongs, in mlx5e_probe.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

CQE compression feature improves performance by reducing PCI bandwidth
bottleneck on CQEs write.
Enhanced CQE compression introduced in ConnectX-6 and it aims to reduce
CPU utilization of SW side packets decompression by eliminating the
need to rewrite ownership bit, which is likely to cost a cache-miss, is
replaced by validity byte handled solely by HW.
Another advantage of the enhanced feature is that session packets are
available to SW as soon as a single CQE slot is filled, instead of
waiting for session to close, this improves packet latency from NIC to
host.

Performance:
Following are tested scenarios and reults comparing basic and enahnced
CQE compression.

setup: IXIA 100GbE connected directly to port 0 and port 1 of
ConnectX-6 Dx 100GbE dual port.

Case #1 RX only, single flow goes to single queue:
IRQ rate reduced by ~ 30%, CPU utilization improved by 2%.

Case #2 IP forwarding from port 1 to port 0 single flow goes to
single queue:
Avg latency improved from 60us to 21us, frame loss improved from 0.5% to 0.0%.

Case #3 IP forwarding from port 1 to port 0 Max Throughput IXIA sends
100%, 8192 UDP flows, goes to 24 queues:
Enhanced is equal or slightly better than basic.

Testing the basic compression feature with this patch shows there is
no perfrormance degradation of the basic compression feature.
Signed-off-by: Ofer Levi <oferle@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>