ixgbe: Fix handling of macvlan Tx offload

This update makes it so that we report the actual number of Tx queues via
real_num_tx_queues but are still restricted to RSS on only the first p

ixgbe: Fix handling of macvlan Tx offload

This update makes it so that we report the actual number of Tx queues via
real_num_tx_queues but are still restricted to RSS on only the first pool
by setting num_tc equal to 1. Doing this locks us into only having the
ability to setup XPS on the queues in that pool, and only those queues
should be used for transmitting anything other than macvlan traffic.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: avoid bringing rings up/down as macvlans are added/removed

This change makes it so that instead of bringing rings up/down for various
we just update the netdev pointer for the Rx

ixgbe: avoid bringing rings up/down as macvlans are added/removed

This change makes it so that instead of bringing rings up/down for various
we just update the netdev pointer for the Rx ring and set or clear the MAC
filter for the interface. By doing it this way we can avoid a number of
races and issues in the code as things were getting messy with the macvlan
clean-up racing with the interface clean-up to bring the rings down on
shutdown.

With this change we opt to leave the rings owned by the PF interface for
both Tx and Rx and just direct the packets once they are received to the
macvlan netdev.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Do not manipulate macvlan Tx queues when performing macvlan offload

We should not be stopping/starting the upper devices Tx queues when
handling a macvlan offload. Instead we shou

ixgbe: Do not manipulate macvlan Tx queues when performing macvlan offload

We should not be stopping/starting the upper devices Tx queues when
handling a macvlan offload. Instead we should be stopping and starting
traffic on our own queues.

In order to prevent us from doing this I am updating the code so that we no
longer change the queue configuration on the upper device, nor do we update
the queue_index on our own device. Instead we can just use the queue index
for our local device and not update the netdev in the case of the transmit
rings.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Don't assume dev->num_tc is equal to hardware TC config

The code throughout ixgbe was assuming that dev->num_tc was populated and
configured with the driver, when in fact this can

ixgbe: Don't assume dev->num_tc is equal to hardware TC config

The code throughout ixgbe was assuming that dev->num_tc was populated and
configured with the driver, when in fact this can be configured via mqprio
without any hardware coordination other than restricting us to the real
number of Tx queues we advertise.

Instead of handling things this way we need to keep a local copy of the
number of TCs in use so that we don't accidentally pull in the TC
configuration from mqprio when it is configured in software mode.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Fix limitations on macvlan so we can support up to 63 offloaded devices

This change is a fix of the macvlan offload so that we correctly handle
macvlan offloaded devices. Specific

ixgbe: Fix limitations on macvlan so we can support up to 63 offloaded devices

This change is a fix of the macvlan offload so that we correctly handle
macvlan offloaded devices. Specifically we were configuring our limits based
on the assumption that we were going to max out the RSS indices for every
mode. As a result when we went to 15 or more macvlan interfaces we were
forced into the 2 queue RSS mode on VFs even though they could have still
supported 4.

This change splits the logic up so that we limit either the total number of
macvlan instances if DCB is enabled, or limit the number of RSS queues used
per macvlan (instead of per pool) if SR-IOV is enabled. By doing this we
can make best use of the part.

In addition I have increased the maximum number of supported interfaces to
63 with one queue per offloaded interface as this more closely reflects the
actual values supported by the interface.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: There is no need to update num_rx_pools in L2 fwd offload

The num_rx_pools value is overwritten when we reinitialize the queue
configuration. In reality we shouldn't need to be up

ixgbe: There is no need to update num_rx_pools in L2 fwd offload

The num_rx_pools value is overwritten when we reinitialize the queue
configuration. In reality we shouldn't need to be updating the value since
it is redone every time we call into ixgbe_setup_tc so for now just drop
the spots where we were incrementing or decrementing the value.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Update adaptive ITR algorithm

The following change is meant to update the adaptive ITR algorithm to
better support the needs of the network. Specifically with this change what

ixgbe: Update adaptive ITR algorithm

The following change is meant to update the adaptive ITR algorithm to
better support the needs of the network. Specifically with this change what
I have done is make it so that our ITR algorithm will try to prevent either
starving a socket buffer for memory in the case of Tx, or overrunning an Rx
socket buffer on receive.

In addition a side effect of the calculations used is that we should
function better with new features such as XDP which can handle small
packets at high rates without needing to lock us into NAPI polling mode.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: NULL xdp_tx rings on resource cleanup

tx_rings and rx_rings are cleaned up on close paths in ixgbe driver
however, xdp_rings are not. Set the xdp_rings to NULL here so that
we

ixgbe: NULL xdp_tx rings on resource cleanup

tx_rings and rx_rings are cleaned up on close paths in ixgbe driver
however, xdp_rings are not. Set the xdp_rings to NULL here so that
we can use the pointer to indicate if the XDP rings are initialized.

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

ixgbe: add support for XDP_TX action

A couple design choices were made here. First I use a new ring
pointer structure xdp_ring[] in the adapter struct instead of
pushing the newly al

ixgbe: add support for XDP_TX action

A couple design choices were made here. First I use a new ring
pointer structure xdp_ring[] in the adapter struct instead of
pushing the newly allocated XDP TX rings into the tx_ring[]
structure. This means we have to duplicate loops around rings
in places we want to initialize both TX rings and XDP rings.
But by making it explicit it is obvious when we are using XDP
rings and when we are using TX rings. Further we don't have
to do ring arithmatic which is error prone. As a proof point
for doing this my first patches used only a single ring structure
and introduced bugs in FCoE code and macvlan code paths.

Second I am aware this is not the most optimized version of
this code possible. I want to get baseline support in using
the most readable format possible and then once this series
is included I will optimize the TX path in another series
of patches.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: get rid of custom busy polling code

In linux-4.5, busy polling was implemented in core
NAPI stack, meaning that all custom implementation can
be removed from drivers.

ixgbe: get rid of custom busy polling code

In linux-4.5, busy polling was implemented in core
NAPI stack, meaning that all custom implementation can
be removed from drivers.

Not only we remove lot's of code, we also remove one lock
operation in fast path, and allow GRO to do its job.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

ixgbe: add mask for 64 RSS queues

The indirection table was reported incorrectly for X550 and newer
where we can support up to 64 RSS queues.

Reported-by Krishneil Singh <krishn

ixgbe: add mask for 64 RSS queues

The indirection table was reported incorrectly for X550 and newer
where we can support up to 64 RSS queues.

Reported-by Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Support 4 queue RSS on VFs with 1 or 2 queue RSS on PF

Instead of limiting the VFs if we don't use 4 queues for RSS in the PF we
can instead just limit the RSS queues used to a po

ixgbe: Support 4 queue RSS on VFs with 1 or 2 queue RSS on PF

Instead of limiting the VFs if we don't use 4 queues for RSS in the PF we
can instead just limit the RSS queues used to a power of 2. By doing this
we can support use cases where VFs are using more queues than the PF is
currently using and can support RSS if so desired.

The only limitation on this is that we cannot support 3 queues of RSS in
the PF or VF. In either of these cases we should fall back to 2 queues in
order to be able to use the power of 2 masking provided by the psrtype
register.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Add support for x550em_a 10G MAC type

Add support for x550em_a 10G MAC type to the ixgbe driver. The new
MAC includes new firmware commands that need to be used to control
PHY

ixgbe: Add support for x550em_a 10G MAC type

Add support for x550em_a 10G MAC type to the ixgbe driver. The new
MAC includes new firmware commands that need to be used to control
PHY and IOSF access, so that support is also added. The interface
supported is a native SFP+ interface.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

net: provide generic busy polling to all NAPI drivers

NAPI drivers no longer need to observe a particular protocol
to benefit from busy polling (CONFIG_NET_RX_BUSY_POLL=y)

napi_

net: provide generic busy polling to all NAPI drivers

NAPI drivers no longer need to observe a particular protocol
to benefit from busy polling (CONFIG_NET_RX_BUSY_POLL=y)

napi_hash_add() and napi_hash_del() are automatically called
from core networking stack, respectively from
netif_napi_add() and netif_napi_del()

This patch depends on free_netdev() and netif_napi_del() being
called from process context, which seems to be the norm.

Drivers might still prefer to call napi_hash_del() on their
own, since they might combine all the rcu grace periods into
a single one, knowing their NAPI structures lifetime, while
core networking stack has no idea of a possible combining.

Once this patch proves to not bring serious regressions,
we will cleanup drivers to either remove napi_hash_del()
or provide appropriate rcu grace periods combining.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

ixgbe: Limit lowest interrupt rate for adaptive interrupt moderation to 12K

This patch updates the lowest limit for adaptive interrupt interrupt
moderation to roughly 12K interrupts per

ixgbe: Limit lowest interrupt rate for adaptive interrupt moderation to 12K

This patch updates the lowest limit for adaptive interrupt interrupt
moderation to roughly 12K interrupts per second.

The way I came about reaching 12K as the desired interrupt rate is by
testing with UDP flows. Specifically I had a simple test that ran a
netperf UDP_STREAM test at varying sizes. What I found was as the packet
sizes increased the performance fell steadily behind until we were only
able to receive at ~4Gb/s with a message size of 65507. A bit of digging
found that we were dropping packets for the socket in the network stack,
and looking at things further what I found was I could solve it by increasing
the interrupt rate, or increasing the rmem_default/rmem_max. What I found was
that when the interrupt coalescing resulted in more data being processed
per interrupt than could be stored in the socket buffer we started losing
packets and the performance dropped. So I reached 12K based on the
following math.

rmem_default = 212992
skb->truesize = 2994
212992 / 2994 = 71.14 packets to fill the buffer

packet rate at 1514 packet size is 812744pps
71.14 / 812744 = 87.9us to fill socket buffer

From there it was just a matter of choosing the interrupt rate and
providing a bit of wiggle room which is why I decided to go with 12K
interrupts per second as that uses a value of 84us.

The data below is based on VM to VM over a direct assigned ixgbe interface.
The test run was:
netperf -H <ip> -t UDP_STREAM"

Socket Message Elapsed Messages CPU Service
Size Size Time Okay Errors Throughput Util Demand
bytes bytes secs # # 10^6bits/sec % SS us/KB
Before:
212992 65507 60.00 1100662 0 9613.4 10.89 0.557
212992 60.00 473474 4135.4 11.27 0.576

After:
212992 65507 60.00 1100413 0 9611.2 10.73 0.549
212992 60.00 974132 8508.3 11.69 0.598

Using bare metal the data is similar but not as dramatic as the throughput
increases from about 8.5Gb/s to 9.5Gb/s.

Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Add new support for X550 MAC's

This patch will add in the new MAC defines and fit it into the switch
cases throughout the driver. New functionality and enablement support will

ixgbe: Add new support for X550 MAC's

This patch will add in the new MAC defines and fit it into the switch
cases throughout the driver. New functionality and enablement support will
be added in following patches.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: add warnings for other disabled features without MSI-X support

When we can't get MSI-X vectors, we disable a few features which require
MSI-X vectors. Print warnings just like we

ixgbe: add warnings for other disabled features without MSI-X support

When we can't get MSI-X vectors, we disable a few features which require
MSI-X vectors. Print warnings just like we do when disabling DCB.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: use e_dev_warn instead of netif_printk

Again, we should not be directly using netif_printk, as we have our own
error print routines that we generate. In addition, instead of using

ixgbe: use e_dev_warn instead of netif_printk

Again, we should not be directly using netif_printk, as we have our own
error print routines that we generate. In addition, instead of using an
early return we can just use the else block of this one line if
statement.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: use e_dev_warn instead of e_err for displaying warning

In this case, disabling DCB is not an error. We can still function, but
we just have to let the user know. In addition, sinc

ixgbe: use e_dev_warn instead of e_err for displaying warning

In this case, disabling DCB is not an error. We can still function, but
we just have to let the user know. In addition, since we call this
during probe before allocating our netdevice structure, we should use
e_dev_warn instead of e_warn.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: determine vector count inside ixgbe_acquire_msix_vectors

Our calculated v_budget doesn't matter except if we allocate MSI-X
vectors. We shouldn't need to calculate this outside of

ixgbe: determine vector count inside ixgbe_acquire_msix_vectors

Our calculated v_budget doesn't matter except if we allocate MSI-X
vectors. We shouldn't need to calculate this outside of the function, so
don't. Instead, only calculate it once we attempt to acquire MSI-X
vectors. This helps collocate all of the MSI-X vector code together.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: move msix_entries allocation into ixgbe_acquire_msix_vectors

We already have to kfree this value if we fail, and this is only part of
MSI-X mode, so we should simply allocate the

ixgbe: move msix_entries allocation into ixgbe_acquire_msix_vectors

We already have to kfree this value if we fail, and this is only part of
MSI-X mode, so we should simply allocate the value where we need it.
This is cleaner, and makes it a lot more obvious why we are freeing it
inside of ixgbe_acquire_msix_vectors.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: return integer from ixgbe_acquire_msix_vectors

Similar to how ixgbevf handles acquiring MSI-X vectors, we can return an
error code instead of relying on the flag being set. This m

ixgbe: return integer from ixgbe_acquire_msix_vectors

Similar to how ixgbevf handles acquiring MSI-X vectors, we can return an
error code instead of relying on the flag being set. This makes it more
clear that we have failed to setup MSI-X mode, and also will make it
easier to consolidate MSI-X related code all into the single function.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: use e_dev_warn instead of netif_printk

The netif_printk relies on our netdevice structure to be registered
already. We may call ixgbe_acquire_msix_vectors prior to registering our

ixgbe: use e_dev_warn instead of netif_printk

The netif_printk relies on our netdevice structure to be registered
already. We may call ixgbe_acquire_msix_vectors prior to registering our
netdevice, so we should not use the netdevice specific printk.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: Refactor busy poll socket code to address multiple issues

This change addresses several issues in the current ixgbe implementation of
busy poll sockets.

First was the fac

ixgbe: Refactor busy poll socket code to address multiple issues

This change addresses several issues in the current ixgbe implementation of
busy poll sockets.

First was the fact that it was possible for frames to be delivered out of
order if they were held in GRO. This is addressed by flushing the GRO buffers
before releasing the q_vector back to the idle state.

The other issue was the fact that we were having to take a spinlock on
changing the state to and from idle. To resolve this I have replaced the
state value with an atomic and use atomic_cmpxchg to change the value from
idle, and a simple atomic set to restore it back to idle after we have
acquired it. This allows us to only use a locked operation on acquiring the
vector without a need for a locked operation to release it.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...

ixgbe: add comment noting recalculation of queues

Since we previously called ixgbe_set_num_queues just prior to attempting
to set our interrupt scheme, it may be non obvious why we have

ixgbe: add comment noting recalculation of queues

Since we previously called ixgbe_set_num_queues just prior to attempting
to set our interrupt scheme, it may be non obvious why we have to call
it again inside the function. Add a comment which helps make it more
obvious that we are resetting features based on the fact that we do not
have MSI-X enabled, and cannot use the previous settings.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

show more ...