>From what I can tell the ConnectX-3 will support an inner IPv6 checksum and
segmentation offload, however it cannot support outer IPv6 headers.  This
assumption is based on the fact that I could see the checksum being
offloaded for inner header on IPv4 tunnels, but not on IPv6 tunnels.

For this reason I am adding the feature to the hw_enc_features and adding
an extra check to the features_check call that will disable GSO and
checksum offload in the case that the encapsulated frame has an outer IP
version of that is not 4.  The check in mlx4_en_features_check could be
removed if at some point in the future a fix is found that allows the
hardware to offload segmentation/checksum on tunnels with an outer IPv6
header.
Signed-off-by: NAlexander Duyck <aduyck@mirantis.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When multiple skb are TX-completed in a row, we might incorrectly keep
a timestamp of a prior skb and cause extra work.

Fixes: ec693d47 ("net/mlx4_en: Add HW timestamping (TS) support")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Reviewed-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Bulk free of SKBs happen transparently by the API call napi_consume_skb().
The napi budget parameter is usually needed by napi_consume_skb()
to detect if called from netpoll.  In this patch it has an extra meaning.

For mlx4 driver, the mlx4_en_stop_port() call is done outside
NAPI/softirq context, and cleanup the entire TX ring via
mlx4_en_free_tx_buf().  The code mlx4_en_free_tx_desc() for
freeing SKBs are shared with NAPI calls.

To handle this shared use the zero budget indication is reused,
and handled appropriately in napi_consume_skb(). To reflect this,
variable is called napi_mode for the function call that needed
this distinction.
Signed-off-by: NJesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

problem description:

The current code sets UAR page size equal to system page size.
The ConnectX-3 and ConnectX-3 Pro HWs require minimum 128 UAR pages.
The mlx4 kernel drivers are not loaded if there is less than 128 UAR pages.

solution:

Always set UAR page to 4KB. This allows more UAR pages if the OS
has PAGE_SIZE larger than 4KB. For example, PowerPC kernel use 64KB
system page size, with 4MB uar region, there are 4MB/2/64KB = 32
uars (half for uar, half for blueflame). This does not meet minimum 128
UAR pages requirement. With 4KB UAR page, there are 4MB/2/4KB = 512 uars
which meet the minimum requirement.

Note that only codes in mlx4_core that deal with firmware know that uar
page size is 4KB. Codes that deal with usr page in cq and qp context
(mlx4_ib, mlx4_en and part of mlx4_core) still have the same assumption
that uar page size equals to system page size.

Note that with this implementation, on 64KB system page size kernel, there
are 16 uars per system page but only one uars is used. The other 15
uars are ignored because of the above assumption.

Regarding SR-IOV, mlx4_core in hypervisor will set the uar page size
to 4KB and mlx4_core code in virtual OS will obtain the uar page size from
firmware.

Regarding backward compatibility in SR-IOV, if hypervisor has this new code,
the virtual OS must be updated. If hypervisor has old code, and the virtual
OS has this new code, the new code will be backward compatible with the
old code. If the uar size is big enough, this new code in VF continues to
work with 64 KB uar page size (on PowerPc kernel). If the uar size does not
meet 128 uars requirement, this new code not loaded in VF and print the same
error message as the old code in Hypervisor.
Signed-off-by: NHuy Nguyen <huyn@mellanox.com>
Reviewed-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We do not set the ins_vlan field to zero when no vlan id is present in the packet.

Since WQEs in the TX ring are not zeroed out between uses, this oversight
could result in having vlan flags present in the WQE ctrl segment when no
vlan is preset.

Fixes: e38af4fa ('net/mlx4_en: Add support for hardware accelerated 802.1ad vlan')
Reported-by: NGideon Naim <gideonn@mellanox.com>
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

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>

Indication of a single completed packet, marked by txbbs_skipped
being bigger then zero, in not enough in order to wake up a
stopped TX queue. The completed packet may contain a single TXBB,
while next packet to be sent (after the wake up) may have multiple
TXBBs (LSO/TSO packets for example), causing overflow in queue followed
by WQE corruption and TX queue timeout.
Instead, wake the stopped queue only when there's enough room for the
worst case (maximum sized WQE) packet that we should need to handle after
the queue is opened again.

Also created an helper routine - mlx4_en_is_tx_ring_full, which checks
if the current TX ring is full or not. It provides better code readability
and removes code duplication.
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>

TX ring QP wasn't released at mlx4_en_destroy_tx_ring. Instead, the code
used the deprecated base_tx_qpn field. Move TX QP release to
mlx4_en_destroy_tx_ring and remove the base_tx_qpn field.

Fixes: ddae0349 ('net/mlx4: Change QP allocation scheme')
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

da91309e (cpumask: Utility function to set n'th cpu...) created a
genuinely weird function.  I never saw it before, it went through DaveM.
(He only does this to make us other maintainers feel better about our own
mistakes.)

cpumask_set_cpu_local_first's purpose is say "I need to spread things
across N online cpus, choose the ones on this numa node first"; you call
it in a loop.

It can fail.  One of the two callers ignores this, the other aborts and
fails the device open.

It can fail in two ways: allocating the off-stack cpumask, or through a
convoluted codepath which AFAICT can only occur if cpu_online_mask
changes.  Which shouldn't happen, because if cpu_online_mask can change
while you call this, it could return a now-offline cpu anyway.

It contains a nonsensical test "!cpumask_of_node(numa_node)".  This was
drawn to my attention by Geert, who said this causes a warning on Sparc.
It sets a single bit in a cpumask instead of returning a cpu number,
because that's what the callers want.

It could be made more efficient by passing the previous cpu rather than
an index, but that would be more invasive to the callers.

Fixes: da91309e
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (then rebased)
Tested-by: NAmir Vadai <amirv@mellanox.com>
Acked-by: NAmir Vadai <amirv@mellanox.com>
Acked-by: NDavid S. Miller <davem@davemloft.net>

By default, the number of tx queues is limited by the number of online cpus
in mlx4_en_get_profile(). However, this limit no longer holds after the
ethtool .set_channels method has been called. In that situation, the driver
may access invalid bits of certain cpumask variables when queue_index >=
nr_cpu_ids.
Signed-off-by: NBenjamin Poirier <bpoirier@suse.de>
Acked-by: NIdo Shamay <idos@mellanox.com>
Fixes: d03a68f8 ("net/mlx4_en: Configure the XPS queue mapping on driver load")
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>

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>

The same macros are used for rx as well. So rename it.
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

iowrite32() will byteswap it's argument on big endian archs.
iowrite32be() will byteswap on little endian archs.
Since we don't want to do this unnecessary byteswap on the fast path,
doorbell is stored in the NIC's native endianness. Using the right
iowrite() according to the arch endianness.

CC: Wei Yang <weiyang@linux.vnet.ibm.com>
CC: David Laight <david.laight@aculab.com>
Fixes: 6a4e8121 ("net/mlx4_en: Avoid calling bswap in tx fast path")
Signed-off-by: NAmir Vadai <amirv@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>

For VXLAN/NVGRE encapsulation, the current HW doesn't support offloading
both the outer UDP TX checksum and the inner TCP/UDP TX checksum.

The driver doesn't advertize SKB_GSO_UDP_TUNNEL_CSUM, however we are wrongly
telling the HW to offload the outer UDP checksum for encapsulated packets,
fix that.

Fixes: 837052d0 ('net/mlx4_en: Add netdev support for TCP/IP
		     offloads of vxlan tunneling')
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

mlx4_en_rx_irq() and mlx4_en_tx_irq() run from hard interrupt context.

They can use napi_schedule_irqoff() instead of napi_schedule()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-By: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add two helpers so that drivers do not have to care of BQL being
available or not.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NJim Davis <jim.epost@gmail.com>
Fixes: 29d40c90 ("net/mlx4_en: Use prefetch in tx path")
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Drivers should avoid NETDEV_TX_BUSY as much as possible.

They should stop the tx queue before qdisc even tries to push another
packet, to avoid requeues.

For a driver supporting skb->xmit_more, this is likely to be a prereq
anyway, otherwise we could have a tx deadlock : We need to force a
doorbell if TX ring is full.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of setting inline threshold using module parameter only on
driver load, use set_tunable() to set it dynamically.
No need to store the threshold per ring, using instead the netdev global
priv->prof->inline_thold
Initial value still is set using the module parameter, therefore
backward compatability is kept.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reorganize code to call is_inline() once, so compiler can inline it
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Properly clear tx_info->ts_requested
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Access skb_headlen() once in tx flow
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Acces skb_shinfo(skb) once in tx flow.
Also, rename @i variable to @i_frag to avoid confusion, as the "goto
tx_drop_unmap;" relied on this @i variable.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

mlx4_en_process_tx_cq() carefully fetches and writes ring->last_nr_txbb
and ring->cons only one time to avoid false sharing
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

mlx4_en_free_tx_desc() uses a prefetchw(&skb->users) to speed up
consume_skb()
prefetchw(&ring->tx_queue->dql) to speed up BQL update
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add frag0_dma/frag0_byte_count into mlx4_en_tx_info to avoid a cache
line miss in TX completion for frames having one dma element.  (We avoid
reading back the tx descriptor)

Note this could be extended to 2/3 dma elements later, as we have free
room in mlx4_en_tx_info

Also, mlx4_en_free_tx_desc() no longer accesses skb_shinfo(). We use a
new nr_maps fields in mlx4_en_tx_info to avoid 2 or 3 cache misses.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Try to allocate using kmalloc_node() first, only on failure use
vmalloc()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

- doorbell_qpn is stored in the cpu_to_be32() way to avoid bswap() in fast
  path.
- mdev->mr.key stored in ring->mr_key to also avoid bswap() and access to
  cold cache line.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

- Remove unused variable ring->poll_cnt
- No need to set some fields if using blueflame
- Add missing const's
- Use unlikely
- Remove unneeded new line
- Make some comments more precise
- struct mlx4_bf @offset field reduced to unsigned int to save space
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ethtool -S reports a new counter, tracking number of time doorbell
was not triggered, because skb->xmit_more was set.

$ ethtool -S eth0 | egrep "tx_packet|xmit_more"
     tx_packets: 2413288400
     xmit_more: 666121277

I merged the tso_packet false sharing avoidance in this patch as well.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

skb->xmit_more tells us if another skb is coming next.

We need to send doorbell when : xmit_more is not set,
or txqueue is stopped (preventing next skb to come immediately)

Tested with a modified pktgen version, I got a 40% increase of
throughput.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This function derives the base address of the CQE from the CQE size,
and calculates the real CQE context segment in it from the factor
(this is like before). Before this change the code used the factor to
calculate the base address of the CQE as well.

The factor indicates in which segment of the cqe stride the cqe information
is located. For 32-byte strides, the segment is 0, and for 64 byte strides,
the segment is 1 (bytes 32..63). Using the factor was ok as long as we had
only 32 and 64 byte strides. However, with larger strides, the factor is zero,
and so cannot be used to calculate the base of the CQE.

The helper uses the same method of CQE buffer pulling made by all other
components that reads the CQE buffer (mlx4_ib driver and libmlx4).
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It would appear the mlx4_en driver was still making a call to
dev_kfree_skb_any() where dev_consume_skb_any() would be more
appropriate.  This should make dropped packet profiling/tracking
easier/better over a NIC driven by mlx4_en.
Signed-off-by: NRick Jones <rick.jones2@hp.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Enable the user to turn off the hardware feature called BlueFlame.
Since it is something specific to mlx4_en hardware, we control
the feature via ethtool private flags.
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It is recommended that TX work not count against the quota.
The cost of TX packet liberation is a minute percentage of what it costs to
process an RX frame. Furthermore, that SKB freeing makes memory available for
other paths in the stack.

Give the TX a larger budget and be more aggressive about cleaning up the Tx
descriptors this budget could be changed using ethtool:
$ ethtool -C eth1 tx-frames-irq <budget>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

IRQ affinity notifier can only have a single notifier - cpu_rmap
notifier. Can't use it to track changes in IRQ affinity map.
Detect IRQ affinity changes by comparing CPU to current IRQ affinity map
during NAPI poll thread.

CC: Thomas Gleixner <tglx@linutronix.de>
CC: Ben Hutchings <ben@decadent.org.uk>
Fixes: 2eacc23c ("net/mlx4_core: Enforce irq affinity changes immediatly")
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Modify the various routines used to allocate memory resources which
serve QPs in mlx4 to get an input GFP directive.  Have the Ethernet
driver to use GFP_KERNEL in it's QP allocations as done prior to this
commit, and the IB driver to use GFP_NOIO when the IB verbs
IB_QP_CREATE_USE_GFP_NOIO QP creation flag is provided.
Signed-off-by: NMel Gorman <mgorman@suse.de>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>