The socket code currently handles link congestion by either blocking
and trying to send again when the congestion has abated, or just
returning to the user with -EAGAIN and let him re-try later.

This mechanism is prone to starvation, because the wakeup algorithm is
non-atomic. During the time the link issues a wakeup signal, until the
socket wakes up and re-attempts sending, other senders may have come
in between and occupied the free buffer space in the link. This in turn
may lead to a socket having to make many send attempts before it is
successful. In extremely loaded systems we have observed latency times
of several seconds before a low-priority socket is able to send out a
message.

In this commit, we simplify this mechanism and reduce the risk of the
described scenario happening. When a message is attempted sent via a
congested link, we now let it be added to the link's backlog queue
anyway, thus permitting an oversubscription of one message per source
socket. We still create a wakeup item and return an error code, hence
instructing the sender to block or stop sending. Only when enough space
has been freed up in the link's backlog queue do we issue a wakeup event
that allows the sender to continue with the next message, if any.

The fact that a socket now can consider a message sent even when the
link returns a congestion code means that the sending socket code can
be simplified. Also, since this is a good opportunity to get rid of the
obsolete 'mtu change' condition in the three socket send functions, we
now choose to refactor those functions completely.
Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

During multicast reception we currently use a simple linked list with
push/pop semantics to store port numbers.

We now see a need for a more generic list for storing values of type
u32. We therefore make some modifications to this list, while replacing
the prefix 'tipc_plist_' with 'u32_'. We also add a couple of new
functions which will come to use in the next commits.
Acked-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The functions tipc_wait_for_sndpkt() and tipc_wait_for_sndmsg() are very
similar. The latter function is also called from two locations, and
there will be more in the coming commits, which will all need to test on
different conditions.

Instead of making yet another duplicates of the function, we now
introduce a new macro tipc_wait_for_cond() where the wakeup condition
can be stated as an argument to the call. This macro replaces all
current and future uses of the two functions, which can now be
eliminated.
Acked-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Sowmini Varadhan says:

====================
TPACKET_V3 TX_RING support

This patch series allows an application to use a single PF_PACKET
descriptor and leverage the best implementations of TX_RING
and RX_RING that exist today.

Patch 1 adds the kernel/Documentation changes for TX_RING
support and patch2 adds the associated test case in selftests.

Changes since v2: additional sanity checks for setsockopt
input for TX_RING/TPACKET_V3. Refactored psock_tpacket.c
test code to avoid code duplication from V2.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a test case and sample code for (TPACKET_V3, PACKET_TX_RING)
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Although TPACKET_V3 Rx has some benefits over TPACKET_V2 Rx, *_v3
does not currently have TX_RING support. As a result an application
that wants the best perf for Tx and Rx (e.g. to handle request/response
transacations) ends up needing 2 sockets, one with *_v2 for Tx and
another with *_v3 for Rx.

This patch enables TPACKET_V2 compatible Tx features in TPACKET_V3
so that an application can use a single descriptor to get the benefits
of _v3 RX_RING and _v2 TX_RING. An application may do a block-send by
first filling up multiple frames in the Tx ring and then triggering a
transmit. This patch only support fixed size Tx frames for TPACKET_V3,
and requires that tp_next_offset must be zero.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

While working with ipmr, we noticed that it is impossible to determine
if an entry is actually unresolved or its IIF interface has disappeared
(e.g. virtual interface got deleted). These entries look almost
identical to user-space when dumping or receiving notifications. So in
order to recognize them add a new RTNH_F_UNRESOLVED flag which is set when
sending an unresolved cache entry to user-space.
Suggested-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some devices, such as the mv88e6097 do have ADDR[0] external and so it
is possible to configure the device to use SMI address 0x1. Remove the
restriction, as there are boards using this address.
Signed-off-by: Volodymyr Bendiuga <volodymyr.bendiuga@westermo.se>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for_4.11/net-next/rds_v3' of git://git.kernel.org/pub/scm/linux/kernel/git/ssantosh/linux

Santosh Shilimkar says:

====================
net: RDS updates

v2->v3:
- Re-based against latest net-next head.
- Dropped a user visible change after discussing with David Miller.
  It needs some more work to fully support old/new tools matrix.
- Addressed Dave's comment about bool usage in patch
  "RDS: IB: track and log active side..."

v1->v2:
Re-aligned indentation in patch 'RDS: mark few internal functions..."

Series consist of:
 - RDMA transport fixes for map failure, listen sequence, handler panic and
   composite message notification.
 - Couple of sparse fixes.
 - Message logging improvements for bind failure, use once mr semantics
   and connection remote address, active end point.
 - Performance improvement for RDMA transport by reducing the post send
   pressure on the queue and spreading the CQ vectors.
 - Useful statistics for socket send/recv usage and receive cache usage.
 - Additional RDS CMSG used by application to track the RDS message
   stages for certain type of traffic to find out latency spots.
   Can be enabled/disabled per socket.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Socket option to tap receive path latency in various stages
in nano seconds. It can be enabled on selective sockets using
using SO_RDS_MSG_RXPATH_LATENCY socket option. RDS will return
the data to application with RDS_CMSG_RXPATH_LATENCY in defined
format. Scope is left to add more trace points for future
without need of change in the interface.
Reviewed-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

RDS support max message size as 1M but the code doesn't check this
in all cases. Patch fixes it for RDMA & non-RDMA and RDS MR size
and its enforced irrespective of underlying transport.
Signed-off-by: Avinash Repaka <avinash.repaka@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Tracks the receive side memory added to scokets and removed from sockets.
Signed-off-by: Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Shutdown code reaping loop takes care of emptying the
CQ's before they being destroyed. And once tasklets are
killed, the hanlders are not expected to run.

But because of core tasklet code issues, tasklet handler could
still run even after tasklet_kill,
RDS IB shutdown code already reaps the CQs before freeing
cq/qp resources so as such the handlers have nothing left
to do post shutdown.

On other hand any handler running after teardown and trying
to access already freed qp/cq resources causes issues
Patch fixes this race by  makes sure that handlers returns
without any action post teardown.
Reviewed-by: Wengang <wen.gang.wang@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

When application sends an RDS RDMA composite message consist of
RDMA transfer to be followed up by non RDMA payload, it expect to
be notified *only* when the full message gets delivered. RDS RDMA
notification doesn't behave this way though.

Thanks to Venkat for debug and root casuing the issue
where only first part of the message(RDMA) was
successfully delivered but remainder payload delivery failed.
In that case, application should not be notified with
a false positive of message delivery success.

Fix this case by making sure the user gets notified only after
the full message delivery.
Reviewed-by: Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Based on available device vectors, allocate cqs accordingly to
get better spread of completion vectors which helps performace
great deal..
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Tracks the ib receive cache total, incoming and frag allocations.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Useful to know the active and passive end points in a
RDS IB connection.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

In absence of extension headers, message log will keep
flooding the console. As such even without use_once we can
clean up the MRs so its not really an error case message
so make it debug message
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

MR invalidation in RDS is done in background thread and not in
data path like registration. So break the dependency between them
which helps to remove the performance bottleneck.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

The first message to a remote node should prompt a new
connection even if it is RDMA operation. For RDMA operation
the MR mapping can fail because connections is not yet up.

Since the connection establishment is asynchronous,
we make sure the map failure because of unavailable
connection reach to the user by appropriate error code.
Before returning to the user, lets trigger the connection
so that its ready for the next retry.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

This prevents RDS from handling incoming rdma packets before RDS
completes initializing its recv/send components.
Signed-off-by: Qing Huang <qing.huang@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Fixes warning: Using plain integer as NULL pointer
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Transport retry is not much useful since it indicate packet loss
in fabric so its better to failover fast rather than longer retry.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Also use pr_* for it.
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

Fixes below warnings:
warning: symbol 'rds_send_probe' was not declared. Should it be static?
warning: symbol 'rds_send_ping' was not declared. Should it be static?
warning: symbol 'rds_tcp_accept_one_path' was not declared. Should it be static?
warning: symbol 'rds_walk_conn_path_info' was not declared. Should it be static?
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

It's useful to know the IP address when RDS fails to bind a
connection. Thus, adding it to the error message.

Orabug: 21894138
Reviewed-by: Wei Lin Guay <wei.lin.guay@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.

The previous implementation of set_settings was modifying
the value of speed and duplex, but with the new API, it's not
possible. The structure ethtool_link_ksettings is defined
as const.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is a function in mdio for the old ethtool api gset.
We add a new function mdio45_ethtool_ksettings_get for the
new ethtool api glinksettings.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
Mellanox mlx5 core and ODP updates 2017-01-01

The following eleven patches mainly come from Artemy Kovalyov
who expanded mlx5 on-demand-paging (ODP) support. In addition
there are three cleanup patches which don't change any functionality,
but are needed to align codebase prior accepting other patches.

Memory region (MR) in IB can be huge and ODP (on-demand paging)
technique allows to use unpinned memory, which can be consumed and
released on demand. This allows to applications do not pin down
the underlying physical pages of the address space, and save from them
need to track the validity of the mappings.

Rather, the HCA requests the latest translations from the OS when pages
are not present, and the OS invalidates translations which are no longer
valid due to either non-present pages or mapping changes.

In existing ODP implementation applications is needed to register
memory buffers for communication, though registered memory regions
need not have valid mappings at registration time.

This patch set performs the following steps to expand
current ODP implementation:

1. It refactors UMR to support large regions, by introducing generic
   function to perform HCA translation table modifications. This
   function supports both atomic and process contexts and is not limited
   by number of modified entries.

   This function allows to enable reallocated memory regions of
   arbitrary size, so adding MR cache buckets to support up to 16GB MRs.

2. It changes page fault event format and refactor page faults logic
   together with addition of atomic support.

3. It prepares mlx5 core code to support implicit registration with
   simplified and relaxed semantics.

   Implicit ODP semantics allows to applications provide special memory
   key that represents their complete address space. Thus all IO accesses
   referencing to this key (with proper access rights associated with the key)
   wouldn't need not register any virtual address range.

Thanks,
        Artemy, Ilya and Leon

v1->v2:
  - Don't use 'inline' in .c files
====================
Signed-off-by: David S. Miller <davem@davemloft.net>