When XDP is active in mlx4, mlx4 is using one page/pkt.
At the same time (i.e. when XDP is active), it is currently
limiting MTU to be FRAG_SZ0 - ETH_HLEN - (2 * VLAN_HLEN)
which is 1514 in x86.  AFAICT, we can at least raise the MTU
limit up to PAGE_SIZE - ETH_HLEN - (2 * VLAN_HLEN) which this
patch is doing.  It will be useful in the next patch which
allows XDP program to extend the packet by adding new header(s).

Note: In the earlier XDP patches, there is already existing guard
to ensure the page/pkt scheme only applies when XDP is active
in mlx4.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While stressing a 40Gbit mlx4 NIC with busy polling, I found false
sharing in mlx4 driver that can be easily avoided.

This patch brings an additional 7 % performance improvement in UDP_RR
workload.

1) If we received no frame during one mlx4_en_process_rx_cq()
   invocation, no need to call mlx4_cq_set_ci() and/or dirty ring->cons

2) Do not refill rx buffers if we have plenty of them.
   This avoids false sharing and allows some bulk/batch optimizations.
   Page allocator and its locks will thank us.

Finally, mlx4_en_poll_rx_cq() should not return 0 if it determined
cpu handling NIC IRQ should be changed. We should return budget-1
instead, to not fool net_rx_action() and its netdev_budget.

v2: keep AVG_PERF_COUNTER(... polled) even if polled is 0
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Do not rearm interrupts if we are busy polling.

mlx4 uses separate CQ for TX and RX, so number of TX interrupts
does not change, unfortunately.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Adam Belay <abelay@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Yuval Mintz <Yuval.Mintz@cavium.com>
Cc: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

XDP statistics are reported in ethtool, in total and per ring,
as follows:
- xdp_drop: the number of packets dropped by xdp.
- xdp_tx: the number of packets forwarded by xdp.
- xdp_tx_full: the number of times an xdp forward failed
	due to a full tx xdp ring.

In addition, all packets that are dropped/forwarded by XDP
are no longer accounted in rx_packets/rx_bytes of the ring,
so that they count traffic that is passed to the stack.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Separately manage the two types of TX rings: regular ones, and XDP.
Upon an XDP set, do not borrow regular TX rings and convert them
into XDP ones, but allocate new ones, unless we hit the max number
of rings.
Which means that in systems with smaller #cores we will not consume
the current TX rings for XDP, while we are still in the num TX limit.

XDP TX rings counters are not shown in ethtool statistics.
Instead, XDP counters will be added to the respective RX rings
in a downstream patch.

This has no performance implications.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use tabs instead of spaces before if statement, no functional change.

Fixes: e7c1c2c4 ("mlx4_en: Added self diagnostics test implementation")
Signed-off-by: NKamal Heib <kamalh@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add likely/unlikely hints to improve branch predictions
in the RX data-path.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Correct drop handling for XDP_TX on TX failure, were recently added in
commit 95357907 ("mlx4: fix XDP_TX is acting like XDP_PASS on TX
ring full").

The change missed an opportunity for recycling the RX page, instead of
going through the page allocator, like the regular XDP_DROP action does.
This patch cease the opportunity, by going through the XDP_DROP case.

Fixes: 95357907 ("mlx4: fix XDP_TX is acting like XDP_PASS on TX ring full")
Signed-off-by: NJesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The XDP_TX action can fail transmitting the frame in case the TX ring
is full or port is down.  In case of TX failure it should drop the
frame, and not as now call 'break' which is the same as XDP_PASS.

Fixes: 9ecc2d86 ("net/mlx4_en: add xdp forwarding and data write support")
Signed-off-by: NJesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: NBrenden Blanco <bblanco@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Depending on the preempt mode, the bpf_prog stored in xdp_prog may be
freed despite the use of call_rcu inside bpf_prog_put. The situation is
possible when running in PREEMPT_RCU=y mode, for instance, since the rcu
callback for destroying the bpf prog can run even during the bh handling
in the mlx4 rx path.

Several options were considered before this patch was settled on:

Add a napi_synchronize loop in mlx4_xdp_set, which would occur after all
of the rings are updated with the new program.
This approach has the disadvantage that as the number of rings
increases, the speed of update will slow down significantly due to
napi_synchronize's msleep(1).

Add a new rcu_head in bpf_prog_aux, to be used by a new bpf_prog_put_bh.
The action of the bpf_prog_put_bh would be to then call bpf_prog_put
later. Those drivers that consume a bpf prog in a bh context (like mlx4)
would then use the bpf_prog_put_bh instead when the ring is up. This has
the problem of complexity, in maintaining proper refcnts and rcu lists,
and would likely be harder to review. In addition, this approach to
freeing must be exclusive with other frees of the bpf prog, for instance
a _bh prog must not be referenced from a prog array that is consumed by
a non-_bh prog.

The placement of rcu_read_lock in this patch is functionally the same as
putting an rcu_read_lock in napi_poll. Actually doing so could be a
potentially controversial change, but would bring the implementation in
line with sk_busy_loop (though of course the nature of those two paths
is substantially different), and would also avoid future copy/paste
problems with future supporters of XDP. Still, this patch does not take
that opinionated option.

Testing was done with kernels in either PREEMPT_RCU=y or
CONFIG_PREEMPT_VOLUNTARY=y+PREEMPT_RCU=n modes, with neither exhibiting
any drawback. With PREEMPT_RCU=n, the extra call to rcu_read_lock did
not show up in the perf report whatsoever, and with PREEMPT_RCU=y the
overhead of rcu_read_lock (according to perf) was the same before/after.
In the rx path, rcu_read_lock is eventually called for every packet
from netif_receive_skb_internal, so the napi poll call's rcu_read_lock
is easily amortized.

v2:
Remove extra rcu_read_lock in mlx4_en_process_rx_cq body
Annotate xdp_prog with __rcu, and convert all usages to rcu_assign or
rcu_dereference[_protected] as appropriate.
Add explicit mutex lock around rcu_assign instead of xchg loop.

Fixes: d576acf0 ("net/mlx4_en: add page recycle to prepare rx ring for tx support")
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Acked-by: NAlexei Starovoitov <alexei.starovoitov@gmail.com>
Signed-off-by: NBrenden Blanco <bblanco@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For consistency, and in order to hint at the synchronous nature of the
xdp_prog field, use READ_ONCE in the destroy path of the ring. All
occurrences should now use either READ_ONCE or xchg.
Signed-off-by: NBrenden Blanco <bblanco@plumgrid.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A user will now be able to loop packets back out of the same port using
a bpf program attached to xdp hook. Updates to the packet contents from
the bpf program is also supported.

For the packet write feature to work, the rx buffers are now mapped as
bidirectional when the page is allocated. This occurs only when the xdp
hook is active.

When the program returns a TX action, enqueue the packet directly to a
dedicated tx ring, so as to avoid completely any locking. This requires
the tx ring to be allocated 1:1 for each rx ring, as well as the tx
completion running in the same softirq.

Upon tx completion, this dedicated tx ring recycles pages without
unmapping directly back to the original rx ring. In steady state tx/drop
workload, effectively 0 page allocs/frees will occur.

In order to separate out the paths between free and recycle, a
free_tx_desc func pointer is introduced that is optionally updated
whenever recycle_ring is activated. By default the original free
function is always initialized.
Signed-off-by: NBrenden Blanco <bblanco@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The mlx4 driver by default allocates order-3 pages for the ring to
consume in multiple fragments. When the device has an xdp program, this
behavior will prevent tx actions since the page must be re-mapped in
TODEVICE mode, which cannot be done if the page is still shared.

Start by making the allocator configurable based on whether xdp is
running, such that order-0 pages are always used and never shared.

Since this will stress the page allocator, add a simple page cache to
each rx ring. Pages in the cache are left dma-mapped, and in drop-only
stress tests the page allocator is eliminated from the perf report.

Note that setting an xdp program will now require the rings to be
reconfigured.

Before:
 26.91%  ksoftirqd/0  [mlx4_en]         [k] mlx4_en_process_rx_cq
 17.88%  ksoftirqd/0  [mlx4_en]         [k] mlx4_en_alloc_frags
  6.00%  ksoftirqd/0  [mlx4_en]         [k] mlx4_en_free_frag
  4.49%  ksoftirqd/0  [kernel.vmlinux]  [k] get_page_from_freelist
  3.21%  swapper      [kernel.vmlinux]  [k] intel_idle
  2.73%  ksoftirqd/0  [kernel.vmlinux]  [k] bpf_map_lookup_elem
  2.57%  swapper      [mlx4_en]         [k] mlx4_en_process_rx_cq

After:
 31.72%  swapper      [kernel.vmlinux]       [k] intel_idle
  8.79%  swapper      [mlx4_en]              [k] mlx4_en_process_rx_cq
  7.54%  swapper      [kernel.vmlinux]       [k] poll_idle
  6.36%  swapper      [mlx4_core]            [k] mlx4_eq_int
  4.21%  swapper      [kernel.vmlinux]       [k] tasklet_action
  4.03%  swapper      [kernel.vmlinux]       [k] cpuidle_enter_state
  3.43%  swapper      [mlx4_en]              [k] mlx4_en_prepare_rx_desc
  2.18%  swapper      [kernel.vmlinux]       [k] native_irq_return_iret
  1.37%  swapper      [kernel.vmlinux]       [k] menu_select
  1.09%  swapper      [kernel.vmlinux]       [k] bpf_map_lookup_elem
Signed-off-by: NBrenden Blanco <bblanco@plumgrid.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add support for the BPF_PROG_TYPE_XDP hook in mlx4 driver.

In tc/socket bpf programs, helpers linearize skb fragments as needed
when the program touches the packet data. However, in the pursuit of
speed, XDP programs will not be allowed to use these slower functions,
especially if it involves allocating an skb.

Therefore, disallow MTU settings that would produce a multi-fragment
packet that XDP programs would fail to access. Future enhancements could
be done to increase the allowable MTU.

The xdp program is present as a per-ring data structure, but as of yet
it is not possible to set at that granularity through any ndo.
Signed-off-by: NBrenden Blanco <bblanco@plumgrid.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Filters cleanup should be done once before destroying net device,
since filters list is contained in the private data.

Fixes: 1eb8c695 ('net/mlx4_en: Add accelerated RFS support')
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use htons instead of unconditionally byte swapping nexthdr.  On a little
endian systems shifting the byte is correct behavior, but it results in
incorrect csums on big endian architectures.

Fixes: f8c6455b ('net/mlx4_en: Extend checksum offloading by CHECKSUM COMPLETE')
Signed-off-by: NDaniel Jurgens <danielj@mellanox.com>
Reviewed-by: NCarol Soto <clsoto@us.ibm.com>
Tested-by: NCarol Soto <clsoto@us.ibm.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The dma_alloc_coherent() function returns a virtual address which can
be used for coherent access to the underlying memory.  On some
architectures, like arm64, undefined behavior results if this memory is
also accessed via virtual mappings that are not coherent.  Because of
their undefined nature, operations like virt_to_page() return garbage
when passed virtual addresses obtained from dma_alloc_coherent().  Any
subsequent mappings via vmap() of the garbage page values are unusable
and result in bad things like bus errors (synchronous aborts in ARM64
speak).

The mlx4 driver contains code that does the equivalent of:
vmap(virt_to_page(dma_alloc_coherent)), this results in an OOPs when the
device is opened.

Prevent Ethernet driver to run this problematic code by forcing it to
allocate contiguous memory. As for the Infiniband driver, at first we
are trying to allocate contiguous memory, but in case of failure roll
back to work with fragmented memory.
Signed-off-by: NHaggai Abramovsky <hagaya@mellanox.com>
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Reported-by: NDavid Daney <david.daney@cavium.com>
Tested-by: NSinan Kaya <okaya@codeaurora.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Count SW packet drops per RX ring instead of a global counter. This
will allow monitoring the number of rx drops per ring.

In addition, SW rx_dropped counter was overwritten by HW rx_dropped
counter, sum both of them instead to show the accurate value.

Fixes: a3333b35 ('net/mlx4_en: Moderate ethtool callback to [...] ')
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Reported-by: NBrenden Blanco <bblanco@plumgrid.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Reported-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes couple error paths after allocation failures.
Atomic set of page reference counter is safe only if it is zero,
otherwise set can race with any speculative get_page_unless_zero.
Signed-off-by: NKonstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

High order pages are optional here since commit 51151a16 ("mlx4: allow
order-0 memory allocations in RX path"), so here is no reason for depleting
reserves. Generic __netdev_alloc_frag() implements the same logic.
Signed-off-by: NKonstantin Khlebnikov <khlebnikov@yandex-team.ru>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The success of CMA allocation largely depends on the success of
migration and key factor of it is page reference count.  Until now, page
reference is manipulated by direct calling atomic functions so we cannot
follow up who and where manipulate it.  Then, it is hard to find actual
reason of CMA allocation failure.  CMA allocation should be guaranteed
to succeed so finding offending place is really important.

In this patch, call sites where page reference is manipulated are
converted to introduced wrapper function.  This is preparation step to
add tracepoint to each page reference manipulation function.  With this
facility, we can easily find reason of CMA allocation failure.  There is
no functional change in this patch.

In addition, this patch also converts reference read sites.  It will
help a second step that renames page._count to something else and
prevents later attempt to direct access to it (Suggested by Andrew).
Signed-off-by: NJoonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Acked-by: NVlastimil Babka <vbabka@suse.cz>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

We would like to automatically provide busy polling support
to all NAPI drivers, without them having to implement anything.

skb_mark_napi_id() can be called from napi_gro_receive() and
napi_get_frags().

Few drivers are still calling skb_mark_napi_id() because
they use netif_receive_skb(). They should eventually call
napi_gro_receive() instead. I will leave this to drivers
maintainers.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Busy polling can now be handled in generic NAPI poll infrastructure.
This removes complexity and fast path overhead :

mlx4 used two spin_lock()/spin_unlock() pair per napi->poll() call
in mlx4_en_cq_lock_napi()/mlx4_en_cq_unlock_napi()

Tested:

Without busy polling :

lpaa23:~# echo 0 >/proc/sys/net/core/busy_read
lpaa24:~# echo 0 >/proc/sys/net/core/busy_read
lpaa23:~# ./netperf -H lpaa24 -t TCP_RR
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to lpaa24.prod.google.com () port 0 AF_INET : first burst 0
Local /Remote
Socket Size   Request  Resp.   Elapsed  Trans.
Send   Recv   Size     Size    Time     Rate
bytes  Bytes  bytes    bytes   secs.    per sec

16384  87380  1        1       10.00    47330.78

With busy polling :

lpaa23:~# echo 70 >/proc/sys/net/core/busy_read
lpaa24:~# echo 70 >/proc/sys/net/core/busy_read
lpaa23:~# ./netperf -H lpaa24 -t TCP_RR
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to lpaa24.prod.google.com () port 0 AF_INET : first burst 0
Local /Remote
Socket Size   Request  Resp.   Elapsed  Trans.
Send   Recv   Size     Size    Time     Rate
bytes  Bytes  bytes    bytes   secs.    per sec

16384  87380  1        1       10.00    97643.55
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When changing rss key, we do not want to overwrite user provided key
by the one provided by netdev_rss_key_fill(), which is the host random
key generated at boot time.

Fixes: 947cbb0a ("net/mlx4_en: Support for configurable RSS hash function")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Eyal Perry <eyalpe@mellanox.com>
CC: Amir Vadai <amirv@mellanox.com>
Acked-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This is a preparatory patch for moving irq_data struct members. Search
and replace was done with coccinelle
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Amir Vadai <amirv@mellanox.com>

To enable device support in accelerated 802.1ad vlan, the port
capability "packet has vlan enable" (phv_en) should be set.
Firmware won't work properly, in case phv_en is not set.

The user can enable "phv_en" port capability with the new ethtool
private flag phv-bit. The phv-bit private flag default value is OFF,
users who are interested in 802.1ad hardware acceleration should turn ON
the phv-bit private flag:
$ ethtool --set-priv-flags eth1 phv-bit on

Once the private flag is set, the device is ready for 802.1ad vlan
acceleration.

The user should also change the interface device features and turn on
"tx-vlan-stag-hw-insert" which is off by default:
$ ethtool -K eth1  tx-vlan-stag-hw-insert on

"phv-bit" private flag setting is available only for Physical
Functions(PF), the Virtual Function (VF) will be able to use the feature
by setting "tx-vlan-stag-hw-insert" ethtool device feature only if the
feature was enabled by the Hypervisor.
Signed-off-by: NHadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To add Hardware accelerated support in 802.1ad vlan, replace
Current VLAN macros to CVLAN.
Replace:
MLX4_WQE_CTRL_INS_VLAN
MLX4_CQE_VLAN_PRESENT_MASK
With:
MLX4_WQE_CTRL_INS_CVLAN
MLX4_CQE_CVLAN_PRESENT_MASK
Signed-off-by: NHadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In mlx4_en_is_ring_empty we check if ring surpassed its size.
Since the prod and cons indicators are u32, there might be a state where
prod wrapped around and cons, making this assert false, although no
actual bug exists (other code segment can cope with this state).
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Mellanox driver has the knowledge if rxhash is a L4 hash,
if it receives a non fragmented TCP or UDP frame and
NETIF_F_RXCSUM is enabled on netdev.

ip_summed value is CHECKSUM_UNNECESSARY in this case.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Amir Vadai <amirv@mellanox.com>
Cc: Ido Shamay <idos@mellanox.com>
Acked-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The check_csum() function relied on hwtstamp_rx_filter to know if rxvlan
offload is disabled. This is wrong since rxvlan offload can be switched
on/off regardless of hwtstamp_rx_filter.

Also moved check_csum to query CQE information to identify VLAN packets
and removed the check of IP packets, since it has been validated before.

Fixes: f8c6455b ('net/mlx4_en: Extend checksum offloading by CHECKSUM COMPLETE')
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Previously, mlx4_en allocated EQs and used them exclusively.
This affected RoCE performance, as applications which are
events sensitive were limited to use only the legacy EQs.

Change that by introducing an EQ pool. This pool is managed
by mlx4_core. EQs are assigned to ports (when there are limited
number of EQs, multiple ports could be assigned to the same EQs).

An exception to this rule is the ASYNC EQ which handles various events.

Legacy EQs are completely removed as all EQs could be shared.

When a consumer (mlx4_ib/mlx4_en) requests an EQ, it asks for
EQ serving on a specific port. The core driver calculates which
EQ should be assigned to that request.

Because IRQs are shared between IB and Ethernet modules, their
names only include the PCI device BDF address.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When system is out of memory, refilling of RX buffers fails while
the driver continue to pass the received packets to the kernel stack.
At some point, when all RX buffers deplete, driver may fall into a
sleep, and not recover when memory for new RX buffers is once again
availible. This is because hardware does not have valid descriptors,
so no interrupt will be generated for the driver to return to work
in napi context. Fix it by schedule the napi poll function from
stats_task delayed workqueue, as long as the allocations fail.
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch should help to improve the performance of the mlx4 and mlx5 on a
number of architectures.  For example, on x86 the dma_wmb/rmb equates out
to a barrer() call as the architecture is already strong ordered, and on
PowerPC the call works out to a lwsync which is significantly less expensive
than the sync call that was being used for wmb.

I placed the new barriers between any spots that seemed to be trying to
order memory/memory reads or writes, if there are any spots that involved
MMIO I left the existing wmb in place as the new barriers cannot order
transactions between coherent and non-coherent memories.

v2: Reduced the replacments to just the spots where I could clearly
    identify the usage pattern.

Cc: Amir Vadai <amirv@mellanox.com>
Cc: Ido Shamay <idos@mellanox.com>
Cc: Eli Cohen <eli@mellanox.com>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Enabled when device supports KEEP FCS. When the flag is set, Ethernet FCS
is appended to the end of the frame, controlled by ethtool.
Signed-off-by: NMuhammad Mahajna <muhammadm@mellanox.com>
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch improves memory utilization and therefore the packets rate
for special MTU's. Instead of setting the frag_stride to the maximal
hard coded frag_size, use the actual frag_size that is set according to
the MTU, when setting the stride of the last frag.
So, for example, for MTU 1600, where the frag_size of the 2nd frag is
86, the frag_size is set to 128 instead of 4096. See below:

Before:
 frag:0 - size:1536 prefix:0 stride:1536
 frag:1 - size:86 prefix:1536 stride:4096

 frag 0 allocator: - size:32768 frags:21
 frag 1 allocator: - size:32768 frags:8

After:
 frag:0 - size:1536 prefix:0 stride:1536
 frag:1 - size:86 prefix:1536 stride:128

 frag 0 allocator: - size:32768 frags:21
 frag 1 allocator: - size:32768 frags:256
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After Initialization of page_alloc, print actual allocated page
size and number of frags it contains. prints is done only when drv
message level is set on the interface.
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Maintain a persistent memory that should survive reset flow/PCI error.
This comes as a preparation for coming series to support above flows.
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.

In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.

When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this  range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.

Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).

Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.

When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.

In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.

The current Ethernet driver code reserves a Tx QP range with 256b alignment.

This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.

This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.

The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:

1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
   and request "BF enabled QPs" if BF is supported for the function

2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0]  - number of QPs
b. param1[31-24] - flags controlling QPs reservation

Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.

Bits 24-30 of the flags are currently reserved.

When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.

In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This was dropped by mistake for the napi_gro_frags flow, fix that.

Fixes: dd65beac ('net/mlx4_en: Extend usage of napi_gro_frags')
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>