Add IP6_TNL_F_USE_ORIG_FWMARK to ip6_tunnel, so that ip6_tnl_xmit2()
makes a route lookup taking into account skb->fwmark and doesnt cache
lookup result.

This permits more flexibility in policies and firewall setups.

To setup such a tunnel, "fwmark inherit" option should be added to "ip
-f inet6 tunnel" command.
Reported-by: NAnders Franzen <Anders.Franzen@ericsson.com>
CC: Hans Schillström <hans.schillstrom@ericsson.com>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit d1b08284 use new frag API but would leave f to be used
uninitialized, this patch fix it.
Signed-off-by: NJason Wang <jasowang@redhat.com>
Acked-by: NMichael S. Tsirkin <mst@redhat.com>
Acked-by: NIan Campbell <ian.campbell@citrix.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of using the multi_tx_table to map possible Tx queues to Tx rings
we can just do simple subtraction for the unlikely event that the Tx queue
provided exceeds the number of Tx rings.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Since igb only uses advanced descriptors we might as well just use an IGB
specific define and drop the _ADV suffix for the descriptor declarations.
In addition this can be further reduced by assuming that it will be working
on pointers since that is normally how the Tx descriptors are handled.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Many of the function names in the hot path are carrying an extra "_adv"
suffix on the end of them to represent the fact that they are using
advanced descriptors instead of legacy descriptors.  However since all igb
uses are advanced descriptors adding the extra suffix doesn't really add
any additional data.  Since this is the case it is easiest to just drop the
suffix and save us from having to store the extra characters.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change is meant to be a general cleanup and performance improvement
for clean_rx_irq.  The previous patch should have updated the allocation so
that the rings can be treated as read-only within the clean_rx_irq
function.  In addition I am re-ordering the operations such that several
goals are accomplished including reducing the overhead for packet
accounting, reducing the number of items on the stack, and improving
overall performance.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change is meant to improve performance by splitting the Tx and Rx
rings into 3 sections.  The first is primarily a read only section
containing basic things like the indexes, a pointer to the dev and netdev
structures, and basic information.  The second section contains the stats
and next_to_use and next_to_clean values.  The third section is primarily
unused values that can just be placed at the end of the ring and are not
used in the hot path.

The adapter structure has several sections that are read in the hot path.
In order to improve performance there I am combining the frequent read
hot path items into a single cache line.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change is meant to streamline the Rx buffer allocation and cleanup.
This is accomplished by reducing the number of writes by only having the Rx
descriptor ring written by software during allocation, and it will only be
read during cleanup.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change removes support for single buffer mode from igb and makes the
driver function in packet split always.  The advantage to doing this is
that we can reduce total memory allocation overhead significantly as we
will only need to allocate one 1K slab per packet and then make use of a
reusable half page instead of allocating a 2K slab per packet.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch modifies the max_frame_size in order account for an optional
VLAN tag.  In order to support this we must also increase the
MAX_STD_JUMBO_FRAME_SIZE to account for the 4 extra bytes.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change cleans up the RXDCTL and TXDCTL configurations and optimizes RX
performance by allowing back write-backs on all hardware other than 82576.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAaron Brown  <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Elimintes prototype link event tracking functionality that has never
been fleshed out and doesn't do anything useful at the current time.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Eliminate the "event_cb" member from TIPC's "subscription" structure
since the function pointer it holds always points to subscr_send_event().
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Modifies the proto_ops structure used by TIPC DGRAM and RDM sockets
so that calls to listen() and accept() are handled by existing kernel
"unsupported operation" routines, and eliminates the related checks
in the listen and accept routines used by SEQPACKET and STREAM sockets
that are no longer needed.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Adds support for the SO_SNDTIMEO socket option. (This complements the
existing support for SO_RCVTIMEO that is already present.)
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Modifies the initial transfer of name table entries to a new neighboring
node so that the messages are enqueued as a unit, rather than individually.

The revised algorithm now locates the link carrying the message only once,
and eliminates unnecessary checks for link congestion, message fragmentation,
and message bundling that are not required when sending these messages.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Functions like this are called using unsigned longs from
function pointers.  In this case, the function is passed in
a node which is normally internally treated as a u32 by TIPC.

Rather than add more casts into this function in the future
for each added use of node within, move the cast to a single
place on a local.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Reduces the maximum size of messages sent during the initial exchange
of name table information between two nodes to be no larger than the
MTU of the first link established between the nodes. This ensures that
messages will never need to be fragmented, which would add unnecessary
overhead to the name table synchronization mechanism.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Reduces the number of bearers a node can support to 2, which can use
identical or non-identical media. This change won't impact users,
since they are currently limited to a maximum of 2 Ethernet bearers,
and will save memory by eliminating a number of unused entries in
TIPC's media and bearer arrays.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Removes obsolete code that searches for an Ethernet bearer structure entry
to use for a newly enabled bearer, since this search is now performed
at the start of the enabling algorithm.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Ensures that the device list lock is held while trying to locate
the Ethernet device used by a newly enabled bearer, so that the
addition or removal of a device does not cause problems.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

Enhances TIPC to ensure that a node that loses contact with a
neighboring node does not allow contact to be re-established until
it sees that its peer has also recognized the loss of contact.

Previously, nodes that were connected by two or more links could
encounter a situation in which node A would lose contact with node B
on all of its links, purge its name table of names published by B,
and then fail to repopulate those names once contact with B was restored.
This would happen because B was able to re-establish one or more links
so quickly that it never reached a point where it had no links to A --
meaning that B never saw a loss of contact with A, and consequently
didn't re-publish its names to A.

This problem is now prevented by enhancing the cleanup done by TIPC
following a loss of contact with a neighboring node to ensure that
node A ignores all messages sent by B until it receives a LINK_PROTOCOL
message that indicates B has lost contact with A, thereby preventing
the (re)establishment of links between the nodes. The loss of contact
is recognized when a RESET or ACTIVATE message is received that has
a "redundant link exists" field of 0, indicating that B's sending link
endpoint is in a reset state and that B has no other working links.

Additionally, TIPC now suppresses the sending of (most) link protocol
messages to a neighboring node while it is cleaning up after an earlier
loss of contact with that node. This stops the peer node from prematurely
activating its link endpoint, which would prevent TIPC from later
activating its own end. TIPC still allows outgoing RESET messages to
occur during cleanup, to avoid problems if its own node recognizes
the loss of contact first and tries to notify the peer of the situation.

Finally, TIPC now recognizes an impending loss of contact with a peer node
as soon as it receives a RESET message on a working link that is the
peer's only link to the node, and ensures that the link protocol
suppression mentioned above goes into effect right away -- that is,
even before its own link endpoints have failed. This is necessary to
ensure correct operation when there are redundant links between the nodes,
since otherwise TIPC would send an ACTIVATE message upon receiving a RESET
on its first link and only begin suppressing when a RESET on its second
link was received, instead of initiating suppression with the first RESET
message as it needs to.

Note: The reworked cleanup code also eliminates a check that prevented
a link endpoint's discovery object from responding to incoming messages
while stale name table entries are being purged. This check is now
unnecessary and would have slowed down re-establishment of communication
between the nodes in some situations.
Signed-off-by: NAllan Stephens <allan.stephens@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

tcp_md5sig_pool is currently an 'array' (a percpu object) of pointers to
struct tcp_md5sig_pool. Only the pointers are NUMA aware, but objects
themselves are all allocated on a single node.

Remove this extra indirection to get proper percpu memory (NUMA aware)
and make code simpler.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Change details:
 - In a continuous sequence of ifconfig up/down operations, there is a small
   window of race between bnad_set_rx_mode() and bnad_cleanup_rx() while the
   former tries to access rx_info->rx & the latter sets it to NULL. This race
   could lead to bna_rx_mode_set() being called with a NULL (rx_info->rx)
   pointer and a crash.
 - Hold bnad->bna_lock while setting / unsetting rx_info->rx in bnad_setup_rx()
   & bnad_cleanup_rx(), thereby eliminating the race described above.
Signed-off-by: NGurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: NRasesh Mody <rmody@brocade.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When Rx queue size is changed, queues are torn down and setup with the new queue
size. During this operation, clear promiscuous mode and restore the original
VLAN filter.
Signed-off-by: NGurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: NRasesh Mody <rmody@brocade.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove a BUG_ON() as it is not required.

Change the unconditional write to release a semaphore to read sem first
and then write. This will eliminate the possibility of sem getting locked
while trying to release it in case if previous sem_get operation failed.
Signed-off-by: NGurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: NRasesh Mody <rmody@brocade.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

netif_tx_start_all_queues() is already called in ixgbe_up_complete, no need
to do it twice.
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Remove duplicate inc of hwstats->ruc
Introduce separate loops for 8 and 16 register reads.
Consolidate mac checks under one case.
Make sure registers are cleared on read.
Reported-by: NJonathan Lynch <jonathan.lynch@thenowfactory.com>
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
CC: Jonathan Lynch <jonathan.lynch@thenowfactory.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch improves the memory utilization with RSC when in one-buffer
mode.  This is accomplished by making the default buffer sizes match up
with the standard memory allocation sizes minus 1K for shared info and
padding overhead.  By doing this CPU utilization when doing large receives
can be reduced by as much as 8%.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The adapter structure was removed from the call so it can be dropped from
the ixgbe_fso documentation.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NRoss Brattain <ross.b.brattain@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

One of the 82598 phys was not being correctly identified as being SFP.
This change corrects that.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change adds a small bit of missing code for enabling the overheat sensor
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe_up and ixgbe_up_complete will always return 0.  Since this doesn't
provide any useful information we might as well just make them both void
and save ourselves from having to return an unused value.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change fixes an issue in which the incorrect amount of headroom was
being reserved for flow director filters.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change fixes a minor redundancy in that tx_sample_rate was set twice.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Private rx_csum flags are now duplicate of netdev->features & NETIF_F_RXCSUM.
Removing this needs deeper surgery.

Things noticed:
 - ixgb has RX csum disabled by default
 - HW VLAN acceleration probably can be toggled, but it's left as is
 - the resets on RX csum offload change can probably be avoided
 - there is A LOT of copy-and-pasted code here
Signed-off-by: NMichał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>