Tom Herbert [Sat, 30 Apr 2016 00:12:15 +0000 (17:12 -0700)]
ipv6: Cleanup IPv6 tunnel receive path
Some basic changes to make IPv6 tunnel receive path look more like
IPv4 path:
- Make ip6_tnl_rcv non-static so that GREv6 and others can call it
- Make ip6_tnl_rcv look like ip_tunnel_rcv
- Switch to gro_cells_receive
- Make ip6_tnl_rcv non-static and export it
Signed-off-by: Tom Herbert <tom@herbertland.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 May 2016 21:02:26 +0000 (17:02 -0400)]
Merge branch 'tcp-preempt'
Eric Dumazet says:
====================
net: make TCP preemptible
Most of TCP stack assumed it was running from BH handler.
This is great for most things, as TCP behavior is very sensitive
to scheduling artifacts.
However, the prequeue and backlog processing are problematic,
as they need to be flushed with BH being blocked.
To cope with modern needs, TCP sockets have big sk_rcvbuf values,
in the order of 16 MB, and soon 32 MB.
This means that backlog can hold thousands of packets, and things
like TCP coalescing or collapsing on this amount of packets can
lead to insane latency spikes, since BH are blocked for too long.
It is time to make UDP/TCP stacks preemptible.
Note that fast path still runs from BH handler.
v2: Added "tcp: make tcp_sendmsg() aware of socket backlog"
to reduce latency problems of large sends.
v3: Fixed a typo in tcp_cdg.c
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:53 +0000 (14:16 -0700)]
tcp: make tcp_sendmsg() aware of socket backlog
Large sendmsg()/write() hold socket lock for the duration of the call,
unless sk->sk_sndbuf limit is hit. This is bad because incoming packets
are parked into socket backlog for a long time.
Critical decisions like fast retransmit might be delayed.
Receivers have to maintain a big out of order queue with additional cpu
overhead, and also possible stalls in TX once windows are full.
Bidirectional flows are particularly hurt since the backlog can become
quite big if the copy from user space triggers IO (page faults)
Some applications learnt to use sendmsg() (or sendmmsg()) with small
chunks to avoid this issue.
Kernel should know better, right ?
Add a generic sk_flush_backlog() helper and use it right
before a new skb is allocated. Typically we put 64KB of payload
per skb (unless MSG_EOR is requested) and checking socket backlog
every 64KB gives good results.
As a matter of fact, tests with TSO/GSO disabled give very nice
results, as we manage to keep a small write queue and smaller
perceived rtt.
Note that sk_flush_backlog() maintains socket ownership,
so is not equivalent to a {release_sock(sk); lock_sock(sk);},
to ensure implicit atomicity rules that sendmsg() was
giving to (possibly buggy) applications.
In this simple implementation, I chose to not call tcp_release_cb(),
but we might consider this later.
Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Alexei Starovoitov <ast@fb.com> Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:52 +0000 (14:16 -0700)]
net: do not block BH while processing socket backlog
Socket backlog processing is a major latency source.
With current TCP socket sk_rcvbuf limits, I have sampled __release_sock()
holding cpu for more than 5 ms, and packets being dropped by the NIC
once ring buffer is filled.
All users are now ready to be called from process context,
we can unblock BH and let interrupts be serviced faster.
cond_resched_softirq() could be removed, as it has no more user.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:51 +0000 (14:16 -0700)]
sctp: prepare for socket backlog behavior change
sctp_inq_push() will soon be called without BH being blocked
when generic socket code flushes the socket backlog.
It is very possible SCTP can be converted to not rely on BH,
but this needs to be done by SCTP experts.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:50 +0000 (14:16 -0700)]
udp: prepare for non BH masking at backlog processing
UDP uses the generic socket backlog code, and this will soon
be changed to not disable BH when protocol is called back.
We need to use appropriate SNMP accessors.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:49 +0000 (14:16 -0700)]
dccp: do not assume DCCP code is non preemptible
DCCP uses the generic backlog code, and this will soon
be changed to not disable BH when protocol is called back.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:48 +0000 (14:16 -0700)]
tcp: do not block bh during prequeue processing
AFAIK, nothing in current TCP stack absolutely wants BH
being disabled once socket is owned by a thread running in
process context.
As mentioned in my prior patch ("tcp: give prequeue mode some care"),
processing a batch of packets might take time, better not block BH
at all.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 Apr 2016 21:16:47 +0000 (14:16 -0700)]
tcp: do not assume TCP code is non preemptible
We want to to make TCP stack preemptible, as draining prequeue
and backlog queues can take lot of time.
Many SNMP updates were assuming that BH (and preemption) was disabled.
Need to convert some __NET_INC_STATS() calls to NET_INC_STATS()
and some __TCP_INC_STATS() to TCP_INC_STATS()
Before using this_cpu_ptr(net->ipv4.tcp_sk) in tcp_v4_send_reset()
and tcp_v4_send_ack(), we add an explicit preempt disabled section.
Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
drivers: net: xgene: Get channel number from device binding
This patch gets ethernet to CPU channel (prefetch buffer number) from
the newly added 'channel' property, thus decoupling Linux driver from
resource management.
Signed-off-by: Iyappan Subramanian <isubramanian@apm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This series adds the driver support for following selftests:
1. Register test
2. Memory test
3. Clock test
4. Interrupt test
5. Internal loopback test
Patch (1) adds the qed driver infrastructure for selftests. Patches (2) and
(3) add qede driver support for ethtool selftests.
Please consider applying this series to "net-next".
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the qede ethtool support for the following tests:
- interrupt test
- memory test
- register test
- clock test
Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com> Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the functionality and APIs needed for selftests.
It adds the ability to configure the link-mode which is required for the
implementation of loopback tests. It adds the APIs for clock test,
register test, interrupt test and memory test.
Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com> Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Andrew Lunn [Fri, 29 Apr 2016 01:24:06 +0000 (21:24 -0400)]
net: dsa: mv88e6xxx: replace ds with ps where possible
The dsa_switch structure ds is actually needed in very few places,
mostly during setup of the switch. The private structure ps is however
needed nearly everywhere. Pass ps, not ds internally.
[vd: rebased Andrew's patch.]
Signed-off-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 May 2016 03:38:49 +0000 (23:38 -0400)]
Merge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue
Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2016-05-01
This series contains updates to i40e and i40evf.
The theme of this series is code reduction, with several code cleanups in
this series. Starting with Neerav's removal of the code that implemented
the HMC AQ APIs and calls, since they are now obsolete and not supported
by firmware.
Anjali changes the default of VFs to make sure they are not trusted or
privileged until its explicitly set for trust through the new NDO op
interface. Also limited the number of MAC and VLAN addresses a VF can
add if it is untrusted/privileged.
Carolyn syncs the VF code for the changes made to the PF for the RSS
hash tuple settings, which ends up cleaning up much of the existing code.
Jesse cleans up compiler warnings which were found with gcc's W=2 option.
Then removed duplicate code, especially since only one copy was actually
being used.
Jacob addresses an issue which was found when testing GCC 6's which
happens to produce new warnings when you left shift a signed value
beyond the storage sizeof the type. The converts i40e & i40evf to use
the BIT() macro more consistently.
Alex actually bucks the trend of code removal by adding support for
both drivers to use GSO_PARTIAL so that segmentation of frames with
checksums enabled in outer headers is supported. Fortunately it does
not take much to add this support!
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp: signal sk_data_ready earlier on data chunks reception
Dave Miller pointed out that fb586f25300f ("sctp: delay calls to
sk_data_ready() as much as possible") may insert latency specially if
the receiving application is running on another CPU and that it would be
better if we signalled as early as possible.
This patch thus basically inverts the logic on fb586f25300f and signals
it as early as possible, similar to what we had before.
Fixes: fb586f25300f ("sctp: delay calls to sk_data_ready() as much as possible") Reported-by: Dave Miller <davem@davemloft.net> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Marek Vasut [Mon, 2 May 2016 00:47:31 +0000 (02:47 +0200)]
mdio_bus: Fix MDIO bus scanning in __mdiobus_register()
Since commit b74766a0a0fe ("phylib: don't return NULL
from get_phy_device()") in linux-next, phy_get_device() will return
ERR_PTR(-ENODEV) instead of NULL if the PHY device ID is all ones.
This causes problem with stmmac driver and likely some other drivers
which call mdiobus_register(). I triggered this bug on SoCFPGA MCVEVK
board with linux-next 20160427 and 20160428. In case of the stmmac, if
there is no PHY node specified in the DT for the stmmac block, the stmmac
driver ( drivers/net/ethernet/stmicro/stmmac/stmmac_mdio.c function
stmmac_mdio_register() ) will call mdiobus_register() , which will
register the MDIO bus and probe for the PHY.
The mdiobus_register() resp. __mdiobus_register() iterates over all of
the addresses on the MDIO bus and calls mdiobus_scan() for each of them,
which invokes get_phy_device(). Before the aforementioned patch, the
mdiobus_scan() would return NULL if no PHY was found on a given address
and mdiobus_register() would continue and try the next PHY address. Now,
mdiobus_scan() returns ERR_PTR(-ENODEV), which is caught by the
'if (IS_ERR(phydev))' condition and the loop exits immediately if the
PHY address does not contain PHY.
Repair this by explicitly checking for the ERR_PTR(-ENODEV) and if this
error comes around, continue with the next PHY address.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: David S. Miller <davem@davemloft.net> Cc: Dinh Nguyen <dinguyen@opensource.altera.com> Cc: Florian Fainelli <f.fainelli@gmail.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Acked-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Alexander Duyck [Thu, 14 Apr 2016 21:19:25 +0000 (17:19 -0400)]
i40e/i40evf: Add support for GSO partial with UDP_TUNNEL_CSUM and GRE_CSUM
This patch makes it so that i40e and i40evf can use GSO_PARTIAL to support
segmentation for frames with checksums enabled in outer headers. As a
result we can now send data over these types of tunnels at over 20Gb/s
versus the 12Gb/s that was previously possible on my system.
The advantage with the i40e parts is that this offload is mostly
transparent as the hardware still deals with the inner and/or outer IPv4
headers so the IP ID is still incrementing for both when this offload is
performed.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Wed, 13 Apr 2016 23:08:27 +0000 (16:08 -0700)]
i40evf: make use of BIT() macro to avoid signed left shift
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Wed, 13 Apr 2016 23:08:26 +0000 (16:08 -0700)]
i40e: make use of BIT() macro to prevent left shift of signed values
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Wed, 13 Apr 2016 23:08:25 +0000 (16:08 -0700)]
i40e/i40evf: fix I40E_MASK signed shift overflow warnings
GCC 6 has a new warning which will display when you attempt to left
shift a signed value beyond the storage size of the type. I40E_MASK
generates a mask value for 32bit registers. Properly typecast the mask
value and place the values in parenthesis to prevent macro expansion
issues.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We have an uninitialized variable warning for valid_len for one case in
validate_vf_mesg. To fix this, just initialize it to 0 at the top of the
function and remove all of the now redundant assignments to 0 in the
individual cases.
Change-Id: Iacbd97f4c521ed8d662eef803a598d8707708cfd Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch syncs the VF code for the changes made to the PF for the RSS
hash tuple settings. Since the VF still cannot change the RSS hash
settings, change the code to make this clear to the user. Previously,
the default settings were returned in this function. However, the
default can be changed by the PF so this does not make sense anymore.
Change-Id: I085eaf005fc7978b440d2a1bf2b2dd7cadaff39b Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
i40e: Limit the number of MAC and VLAN addresses that can be added for VFs
If the VF is privileged/trusted it can do as it may please including
but not limited to hogging resources and playing unfair.
But if the VF is not privileged/trusted it still can add some number
(8) of MAC and VLAN addresses.
Other restrictions with respect to Port VLAN and normal VLAN still apply
to not privileged/trusted VF.
Change-Id: I3a9529201b184c8873e1ad2e300aff468c9e6296 Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jon Paul Maloy [Fri, 29 Apr 2016 00:16:08 +0000 (20:16 -0400)]
tipc: set 'active' state correctly for first established link
When we are displaying statistics for the first link established between
two peers, it will always be presented as STANDBY although it in reality
is ACTIVE.
This happens because we forget to set the 'active' flag in the link
instance at the moment it is established. Although this is a bug, it only
has impact on the presentation view of the link, not on its actual
functionality.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
of: of_mdio: Check if MDIO bus controller is available
Add a check whether the 'struct device_node' pointer passed to
of_mdiobus_register() is an available (aka enabled) node in the Device
Tree.
Rationale for doing this are cases where an Ethernet MAC provides a MDIO
bus controller and node, and an additional Ethernet MAC might be
connecting its PHY/switches to that first MDIO bus controller, while
still embedding one internally which is therefore marked as "disabled".
Instead of sprinkling checks like these in callers of
of_mdiobus_register(), do this in a central location.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 20:29:12 +0000 (16:29 -0400)]
Merge branch 'mlx5-aRFS'
Saeed Mahameed says:
====================
Mellanox 100G mlx5 ethernet aRFS support
This series adds accelerated RFS support for the mlx5e driver.
I have added one patch non-related to aRFS that fixes the rtnl_lock
warning mlx5 driver been getting since b7aade15485a ('vxlan: break dependency with netdev drivers')
aRFS support in details:
A direct TIR per RQ is now required in order to have the essential building blocks
for aRFS. Today the driver has one direct TIR that forwards traffic to RQ[0] (core 0),
and one indirect TIR for RSS indirection table. For that we've added one direct TIR
per RQ, e.g.: TIR[i] -> RQ[i] (core i).
Publicize Modify flow rule destination and reveal it in flow steering API, to have the
ability to dynamically modify the destination TIR(core) for aRFS rules from the
ethernet driver.
Initializing CPU reverse mapping to notify upper layer on internal receive queue cpu
mappings.
Some design refactoring for mlx5e ethernet driver flow tables and flow steering API.
Now the caller of create_flow_table can choose the level of the flow table, this way
we will create the mlx5e flow tables in a reversed order and connect them as we go,
we create flow table[i+1] before flow table[i] to be able to set flow table[i + 1] as
a destination of flow table[i] once flow table[i] is created.
also we have split the main flow table in the following manner:
- From before: RX packet had to visit two flow tables until it is delivered to its receive queue:
RX packet -> vlan filter flow table -> main flow table.
> vlan filter will check the packet vlan field is allowed.
> main flow will check if the dest mac is allowed and will check the l3/l4 headers to
retrieve the RSS hash for steering the packet into its final receive queue.
- Now main flow table is split into l2 dst mac steering table and ttc (traffic type classifier) table:
RX packet -> vlan filter -> l2 table -> ttc table
> vlan filter - same as before
> L2 filter - filter packets according their destination mac address
> ttc table - classify packet headers for RSS steering
- L3/L4 classification rules to steer the packet according to thier headers hash
- in case of none of the rules applies the packet is steered to RQ[0]
After the above refactoring all left to-do is to create aRFS flow table which will manage
aRFS steering rules to forward traffic to the desired RQ (core) and just connect the ttc
table rules destinations to aRFS flow table.
aRFS flow table in case of a miss will deliver the traffic to the core where the original
ttc hash would have chosen.
TTC table is not initialized and enabled until the user explicitly asks to, i.e. setting the NETIF_F_NTUPLE
to ON. This way there is no need for ttc table to forward traffic to aRFS table unless required.
When setting back to OFF aRFS flow table is disabled and disconnected.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:42 +0000 (01:36 +0300)]
net/mlx5e: Enabling aRFS mechanism
Accelerated RFS requires that ntuple filtering is enabled via
ethtool and driver supports ndo_rx_flow_steer.
When the ntuple filtering is enabled, we modify the l3_l4 ttc
rules to point on the aRFS flow tables and when the filtering
is disabled, we modify the l3_l4 ttc rules to point on the RSS
TIRs.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:41 +0000 (01:36 +0300)]
net/mlx5e: Add accelerated RFS support
Implement ndo_rx_flow_steer ndo.
A new flow steering rule will be composed from the
skb 4-tuple and added to the hardware aRFS flow table.
Each rule is stored in an internal hash table, if such
skb 4-tuple rule already exists we update the corresponding
hardware steering rule with the new destination.
For garbage collection rps_may_expire_flow will be
invoked for a limited amount of old rules upon any
ndo_rx_flow_steer invocation.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:40 +0000 (01:36 +0300)]
net/mlx5e: Create aRFS flow tables
Create the following four flow tables for aRFS usage:
1. IPv4 TCP - filtering 4-tuple of IPv4 TCP packets.
2. IPv6 TCP - filtering 4-tuple of IPv6 TCP packets.
3. IPv4 UDP - filtering 4-tuple of IPv4 UDP packets.
4. IPv6 UDP - filtering 4-tuple of IPv6 UDP packets.
Each flow table has two flow groups: one for the 4-tuple
filtering (full match) and the other contains * rule for miss rule.
Full match rule means a hit for aRFS and packet will be forwarded
to the dedicated RQ/Core, miss rule packets will be forwarded to
default RSS hashing.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:39 +0000 (01:36 +0300)]
net/mlx5: Initializing CPU reverse mapping
Allocating CPU rmap and add entry for each IRQ.
CPU rmap is used in aRFS to get the RX queue number
of the RX completion interrupts.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:38 +0000 (01:36 +0300)]
net/mlx5e: Split the main flow steering table
Currently, the main flow table is used for two purposes:
One is to do mac filtering and the other is to classify
the packet l3-l4 header in order to steer the packet to
the right RSS TIR.
This design is very complex, for each configured mac address we
have to add eleven rules (rule for each traffic type), the same if the
device is put to promiscuous/allmulti mode.
This scheme isn't scalable for future features like aRFS.
In order to simplify it, the main flow table is split to two flow
tables:
1. l2 table - filter the packet dmac address, if there is a match
we forward to the ttc flow table.
2. TTC (Traffic Type Classifier) table - classify the traffic
type of the packet and steer the packet to the right TIR.
In this new design, when new mac address is added, the driver adds
only one flow rule instead of eleven.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:37 +0000 (01:36 +0300)]
net/mlx5e: Refactor mlx5e flow steering structs
Slightly refactor and re-order the flow steering structs,
tables and data-bases for better self-containment and
flexibility to add more future steering phases
(tables/rules/data bases) e.g: aRFS.
Changes:
1. Move the vlan DB and address DB into their table structs.
2. Rename steering table structs to unique format: mlx5e_*_table,
e.g: mlx5e_vlan_table.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:36 +0000 (01:36 +0300)]
net/mlx5: Support different attributes for priorities in namespace
Currently, namespace could be initialized only
with priorities with the same attributes.
Add support to initialize namespace with priorities
with different attributes(e.g. different number of levels).
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:35 +0000 (01:36 +0300)]
net/mlx5: Add user chosen levels when allocating flow tables
Currently, consumers of the flow steering infrastructure can't
choose their own flow table levels and are limited to one
flow table per level. This just waste levels.
Instead, we introduce here the possibility to use multiple
flow tables in a level. The user is free to connect these
flow tables, while following the rule (FTEs in FT of level x
could only point to FTs of level y where y > x).
In addition this patch switch the order of the create/destroy
flow tables of the NIC(vlan and main).
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:34 +0000 (01:36 +0300)]
net/mlx5: Set number of allowed levels in priority
Refactors the flow steering namespace creation,
by changing the name num_fts to num_levels.
When new flow table is created, the driver assign new level
to this flow table therefore the meaning is equivalent.
Since downstream patches will introduce the ability to create more
than one flow table per level, the name num_fts is no
longer accurate.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Maor Gottlieb [Thu, 28 Apr 2016 22:36:33 +0000 (01:36 +0300)]
net/mlx5: Introduce modify flow rule destination
This API is used for modifying the flow rule destination.
This is needed for modifying the pointed flow table by the
traffic type classifier rules to point on the aRFS tables.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce new TIRs for direct access per RQ.
Now we have 2 available kinds of TIRs:
- indirect TIR per traffic type, each points to one RQT (RSS RQT)
same as before.
- New direct TIR per RQ, each points to RQT with a size of one
that forwards packets to that RQ only.
Driver will open max channels (num cores) direct TIRs by default,
they will be filled with the actual RQs once channels are allocated.
Needed for downstream aRFS and ethtool direct steering functionalities.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Matthew Finlay [Thu, 28 Apr 2016 22:36:31 +0000 (01:36 +0300)]
net/mlx5e: Call vxlan_get_rx_port() with rtnl lock
Hold the rtnl lock when calling vxlan_get_rx_port().
Fixes: b7aade15485a ("vxlan: break dependency with netdev drivers") Signed-off-by: Matthew Finlay <matt@mellanox.com> Reported-by: Alexander Duyck <alexander.duyck@gmail.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 20:23:03 +0000 (16:23 -0400)]
Merge branch 'enc28j60-small-improvements'
Michael Heimpold says:
====================
net: ethernet: enc28j60: small improvements
This series of two patches adds the following improvements to the driver:
1) Rework the central SPI read function so that it is compatible with
SPI masters which only support half duplex transfers.
2) Add a device tree binding for the driver.
Changelog:
v3: * renamed and improved binding documentation as
suggested by Rob Herring
v2: * took care of Arnd Bergmann's review comments
- allow to specify MAC address via DT
- unconditionally define DT id table
* increased the driver version minor number
* driver author's email address bounces, removed from address list
v1: * Initial submission
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Michael Heimpold [Thu, 28 Apr 2016 20:06:15 +0000 (22:06 +0200)]
net: ethernet: enc28j60: add device tree support
The following patch adds the required match table for device tree support
(and while at, fix the indent). It's also possible to specify the
MAC address in the DT blob.
Also add the corresponding binding documentation file.
Signed-off-by: Michael Heimpold <mhei@heimpold.de> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 20:09:45 +0000 (16:09 -0400)]
Merge branch 'ppp-rtnetlink'
Guillaume Nault says:
====================
ppp: add rtnetlink support
PPP devices lack the ability to be customised at creation time. In
particular they can't be created in a given netns or with a particular
name. Moving or renaming the device after creation is possible, but
creates undesirable transient effects on servers where PPP devices are
constantly created and removed, as users connect and disconnect.
Implementing rtnetlink support solves this problem.
The rtnetlink handlers implemented in this series are minimal, and can
only replace the PPPIOCNEWUNIT ioctl. The rest of PPP ioctls remains
necessary for any other operation on channels and units.
It is perfectly possible to mix PPP devices created by rtnl
and by ioctl(PPPIOCNEWUNIT). Devices will behave in the same way.
mutex_trylock() is used to resolve the locking issue wrt. locking
dependency between rtnl_lock() and ppp_mutex (see ppp_nl_newlink() in
patch #2).
A user visible difference brought by this series is that old PPP
interfaces (those created with ioctl(PPPIOCNEWUNIT)), can now be
removed by "ip link del", just like new rtnl based PPP devices.
Changes since v3:
- Rebase on net-next.
- Not an RFC anymore.
Changes since v2:
- Define ->rtnl_link_ops for ioctl based PPP devices, so they can
handle rtnl messages just like rtnl based ones (suggested by
Stephen Hemminger).
- Move back to original lock ordering between ppp_mutex and rtnl_lock
to simplify patch series. Handle lock inversion issue using
mutex_trylock() (suggested by Stephen Hemminger).
- Do file descriptor lookup directly in ppp_nl_newlink(), to simplify
ppp_dev_configure().
Changes since v1:
- Rebase on net-next.
- Invert locking order wrt. ppp_mutex and rtnl_lock and protect
file->private_data with ppp_mutex.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Define PPP device handler for use with rtnetlink.
The only PPP specific attribute is IFLA_PPP_DEV_FD. It is mandatory and
contains the file descriptor of the associated /dev/ppp instance (the
file descriptor which would have been used for ioctl(PPPIOCNEWUNIT) in
the ioctl-based API). The PPP device is removed when this file
descriptor is released (same behaviour as with ioctl based PPP
devices).
PPP devices created with the rtnetlink API behave like the ones created
with ioctl(PPPIOCNEWUNIT). In particular existing ioctls work the same
way, no matter how the PPP device was created.
The rtnl callbacks are also assigned to ioctl based PPP devices. This
way, rtnl messages have the same effect on any PPP devices.
The immediate effect is that all PPP devices, even ioctl-based
ones, can now be removed with "ip link del".
A minor difference still exists between ioctl and rtnl based PPP
interfaces: in the device name, the number following the "ppp" prefix
corresponds to the PPP unit number for ioctl based devices, while it is
just an unrelated incrementing index for rtnl ones.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr> Signed-off-by: David S. Miller <davem@davemloft.net>
vxlan: fix initialization with custom link parameters
Commit 0c867c9bf84c ("vxlan: move Ethernet initialization to a separate
function") changed initialization order and as an unintended result, when the
user specifies additional link parameters (such as IFLA_ADDRESS) while
creating vxlan interface, those are overwritten by vxlan_ether_setup later.
It's necessary to call ether_setup from withing the ->setup callback. That
way, the correct parameters are set by rtnl_create_link later. This is done
also for VXLAN-GPE, as we don't know the interface type yet at that point,
and changed to the correct interface type later.
Fixes: 0c867c9bf84c ("vxlan: move Ethernet initialization to a separate function") Reported-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Signed-off-by: Jiri Benc <jbenc@redhat.com> Tested-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 18:26:32 +0000 (14:26 -0400)]
Merge branch 'samples-bpf-user-experience'
Jesper Dangaard Brouer says:
====================
samples/bpf: Improve user experience
It is a steep learning curve getting started with using the eBPF
examples in samples/bpf/. There are several dependencies, and
specific versions of these dependencies. Invoking make in the correct
manor is also slightly obscure.
This patchset cleanup, document and hopefully improves the first time
user experience with the eBPF samples directory by auto-detecting
certain scenarios.
V4:
- Address Naveen's nitpicks
- Handle/fail if extra args are passed in LLC or CLANG (David Laight)
V3:
- Add Alexei's ACKs
- Remove README paragraph about LLVM experimental BPF target
as it only existed between LLVM version 3.6 to 3.7.
V2:
- Adjusted recommend minimum versions to 3.7.1
- Included clang build instructions
- New patch adding CLANG variable and validation of command
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
samples/bpf: like LLC also verify and allow redefining CLANG command
Users are likely to manually compile both LLVM 'llc' and 'clang'
tools. Thus, also allow redefining CLANG and verify command exist.
Makefile implementation wise, the target that verify the command have
been generalized.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
samples/bpf: allow make to be run from samples/bpf/ directory
It is not intuitive that 'make' must be run from the top level
directory with argument "samples/bpf/" to compile these eBPF samples.
Introduce a kbuild make file trick that allow make to be run from the
"samples/bpf/" directory itself. It basically change to the top level
directory and call "make samples/bpf/" with the "/" slash after the
directory name.
Also add a clean target that only cleans this directory, by taking
advantage of the kbuild external module setting M=$PWD.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
samples/bpf: add a README file to get users started
Getting started with using examples in samples/bpf/ is not
straightforward. There are several dependencies, and specific
versions of these dependencies.
Just compiling the example tool is also slightly obscure, e.g. one
need to call make like:
make samples/bpf/
Do notice the "/" slash after the directory name.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
samples/bpf: Makefile verify LLVM compiler avail and bpf target is supported
Make compiling samples/bpf more user friendly, by detecting if LLVM
compiler tool 'llc' is available, and also detect if the 'bpf' target
is available in this version of LLVM.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
samples/bpf: add back functionality to redefine LLC command
It is practical to be-able-to redefine the location of the LLVM
command 'llc', because not all distros have a LLVM version with bpf
target support. Thus, it is sometimes required to compile LLVM from
source, and sometimes it is not desired to overwrite the distros
default LLVM version.
This feature was removed with 128d1514be35 ("samples/bpf: Use llc in
PATH, rather than a hardcoded value").
Add this features back. Note that it is possible to redefine the LLC
on the make command like:
make samples/bpf/ LLC=~/git/llvm/build/bin/llc
Fixes: 128d1514be35 ("samples/bpf: Use llc in PATH, rather than a hardcoded value") Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 17:41:47 +0000 (13:41 -0400)]
Merge branch 'cxgb4-mbox-cmd-logging'
Hariprasad Shenai says:
====================
cxgb4/cxgb4vf: add support for mbox cmd logging
This patch series adds support for logging mailbox commands and
replies for debugging purpose for both PF and VF driver.
This patch series has been created against net-next tree and includes
patches on cxgb4 and cxgb4vf driver.
We have included all the maintainers of respective drivers. Kindly
review the change and let us know in case of any review comments.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 Apr 2016 17:39:04 +0000 (13:39 -0400)]
Merge branch 'hns-props'
Yisen Zhuang says:
====================
net: hns: update DT properties according to Rob's comments
There are some inappropriate properties definition in hns DT. We
update the definition according to Rob's review comments and fix some
typos in binding.
For more details, please see individual patches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
ipvlan: Fix failure path in dev registration during link creation
When newlink creation fails at device-registration, the port->count
is decremented twice. Francesco Ruggeri (fruggeri@arista.com) found
this issue in Macvlan and the same exists in IPvlan driver too.
While fixing this issue I noticed another issue of missing unregister
in case of failure, so adding it to the fix which is similar to the
macvlan fix by Francesco in commit 308379607548 ("macvlan: fix failure
during registration v3")
Reported-by: Francesco Ruggeri <fruggeri@arista.com> Signed-off-by: Mahesh Bandewar <maheshb@google.com> CC: Eric Dumazet <edumazet@google.com> CC: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: David S. Miller <davem@davemloft.net>
françois romieu [Wed, 27 Apr 2016 21:29:44 +0000 (23:29 +0200)]
pch_gbe: replace private tx ring lock with common netif_tx_lock
pch_gbe_tx_ring.tx_lock is only used in the hard_xmit handler and
in the transmit completion reaper called from NAPI context.
Compile-tested only. Potential victims Cced.
Someone more knowledgeable may check if pch_gbe_tx_queue could
have some use for a mmiowb.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com> Cc: Darren Hart <dvhart@infradead.org> Cc: Andy Cress <andy.cress@us.kontron.com> Cc: bryan@fossetcon.org Signed-off-by: David S. Miller <davem@davemloft.net>
net: dsa: Provide CPU port statistics to master netdev
This patch overloads the DSA master netdev, aka CPU Ethernet MAC to also
include switch-side statistics, which is useful for debugging purposes,
when the switch is not properly connected to the Ethernet MAC (duplex
mismatch, (RG)MII electrical issues etc.).
We accomplish this by retaining the original copy of the master netdev's
ethtool_ops, and just overload the 3 operations we care about:
get_sset_count, get_strings and get_ethtool_stats so as to intercept
these calls and call into the original master_netdev ethtool_ops, plus
our own.
We take this approach as opposed to providing a set of DSA helper
functions that would retrive the CPU port's statistics, because the
entire purpose of DSA is to allow unmodified Ethernet MAC drivers to be
used as CPU conduit interfaces, therefore, statistics overlay in such
drivers would simply not scale.
The new ethtool -S <iface> output would therefore look like this now:
<iface> statistics
p<2 digits cpu port number>_<switch MIB counter names>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 27 Apr 2016 17:12:25 +0000 (10:12 -0700)]
tcp: give prequeue mode some care
TCP prequeue goal is to defer processing of incoming packets
to user space thread currently blocked in a recvmsg() system call.
Intent is to spend less time processing these packets on behalf
of softirq handler, as softirq handler is unfair to normal process
scheduler decisions, as it might interrupt threads that do not
even use networking.
Current prequeue implementation has following issues :
1) It only checks size of the prequeue against sk_rcvbuf
It was fine 15 years ago when sk_rcvbuf was in the 64KB vicinity.
But we now have ~8MB values to cope with modern networking needs.
We have to add sk_rmem_alloc in the equation, since out of order
packets can definitely use up to sk_rcvbuf memory themselves.
2) Even with a fixed memory truesize check, prequeue can be filled
by thousands of packets. When prequeue needs to be flushed, either
from sofirq context (in tcp_prequeue() or timer code), or process
context (in tcp_prequeue_process()), this adds a latency spike
which is often not desirable.
I added a fixed limit of 32 packets, as this translated to a max
flush time of 60 us on my test hosts.
Also note that all packets in prequeue are not accounted for tcp_mem,
since they are not charged against sk_forward_alloc at this point.
This is probably not a big deal.
Note that this might increase LINUX_MIB_TCPPREQUEUEDROPPED counts,
which is misnamed, as packets are not dropped at all, but rather pushed
to the stack (where they can be either consumed or dropped)
Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Michal Kazior [Wed, 27 Apr 2016 10:59:13 +0000 (12:59 +0200)]
fq: split out backlog update logic
mac80211 (which will be the first user of the
fq.h) recently started to support software A-MSDU
aggregation. It glues skbuffs together into a
single one so the backlog accounting needs to be
more fine-grained.
To avoid backlog sorting logic duplication split
it up for re-use.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Dan Carpenter [Wed, 27 Apr 2016 08:05:28 +0000 (11:05 +0300)]
tipc: remove an unnecessary NULL check
This is never called with a NULL "buf" and anyway, we dereference 's' on
the lines before so it would Oops before we reach the check.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net/mlx5e: avoid stack overflow in mlx5e_open_channels
struct mlx5e_channel_param is a large structure that is allocated
on the stack of mlx5e_open_channels, and with a recent change
it has grown beyond the warning size for the maximum stack
that a single function should use:
mellanox/mlx5/core/en_main.c: In function 'mlx5e_open_channels':
mellanox/mlx5/core/en_main.c:1325:1: error: the frame size of 1072 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
The function is already using dynamic allocation and is not in
a fast path, so the easiest workaround is to use another kzalloc
for allocating the channel parameters.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Fixes: d3c9bc2743dc ("net/mlx5e: Added ICO SQs") Acked-by: Saeed Mahameed <saeedm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Jason Wang [Tue, 26 Apr 2016 03:13:42 +0000 (23:13 -0400)]
tuntap: calculate rps hash only when needed
There's no need to calculate rps hash if it was not enabled. So this
patch export rps_needed and check it before trying to get rps
hash. Tests (using pktgen to inject packets to guest) shows this can
improve pps about 13% (when rps is disabled).
Cc: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Jason Wang <jasowang@redhat.com>
----
Changes from V1:
- Fix build when CONFIG_RPS is not set Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 Apr 2016 20:14:20 +0000 (16:14 -0400)]
Merge branch 'tcp-eor'
Martin KaFai Lau says:
====================
tcp: Make use of MSG_EOR in tcp_sendmsg
v4:
~ Do not set eor bit in do_tcp_sendpages() since there is
no way to pass MSG_EOR from the userland now.
~ Avoid rmw by testing MSG_EOR first in tcp_sendmsg().
~ Move TCP_SKB_CB(skb)->eor test to a new helper
tcp_skb_can_collapse_to() (suggested by Soheil).
~ Add some packetdrill tests.
v3:
~ Separate EOR marking from the SKBTX_ANY_TSTAMP logic.
~ Move the eor bit test back to the loop in tcp_sendmsg and
tcp_sendpage because there could be >1 threads doing
sendmsg.
~ Thanks to Eric Dumazet's suggestions on v2.
~ The TCP timestamp bug fixes are separated into other threads.
v2:
~ Rework based on the recent work
"add TX timestamping via cmsg" by
Soheil Hassas Yeganeh <soheil.kdev@gmail.com>
~ This version takes the MSG_EOR bit as a signal of
end-of-response-message and leave the selective
timestamping job to the cmsg
~ Changes based on the v1 feedback (like avoid
unlikely check in a loop and adding tcp_sendpage
support)
~ The first 3 patches are bug fixes. The fixes in this
series depend on the newly introduced txstamp_ack in
net-next. I will make relevant patches against net after
getting some feedback.
~ The test results are based on the recently posted net fix:
"tcp: Fix SOF_TIMESTAMPING_TX_ACK when handling dup acks"
One potential use case is to use MSG_EOR with
SOF_TIMESTAMPING_TX_ACK to get a more accurate
TCP ack timestamping on application protocol with
multiple outgoing response messages (e.g. HTTP2).
One of our use case is at the webserver. The webserver tracks
the HTTP2 response latency by measuring when the webserver sends
the first byte to the socket till the TCP ACK of the last byte
is received. In the cases where we don't have client side
measurement, measuring from the server side is the only option.
In the cases we have the client side measurement, the server side
data can also be used to justify/cross-check-with the client
side data.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Martin KaFai Lau <kafai@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Soheil Hassas Yeganeh <soheil@google.com> Cc: Willem de Bruijn <willemb@google.com> Cc: Yuchung Cheng <ycheng@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Mon, 25 Apr 2016 21:44:49 +0000 (14:44 -0700)]
tcp: Handle eor bit when coalescing skb
This patch:
1. Prevent next_skb from coalescing to the prev_skb if
TCP_SKB_CB(prev_skb)->eor is set
2. Update the TCP_SKB_CB(prev_skb)->eor if coalescing is
allowed
Signed-off-by: Martin KaFai Lau <kafai@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Soheil Hassas Yeganeh <soheil@google.com> Cc: Willem de Bruijn <willemb@google.com> Cc: Yuchung Cheng <ycheng@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Mon, 25 Apr 2016 21:44:48 +0000 (14:44 -0700)]
tcp: Make use of MSG_EOR in tcp_sendmsg
This patch adds an eor bit to the TCP_SKB_CB. When MSG_EOR
is passed to tcp_sendmsg, the eor bit will be set at the skb
containing the last byte of the userland's msg. The eor bit
will prevent data from appending to that skb in the future.
The change in do_tcp_sendpages is to honor the eor set
during the previous tcp_sendmsg(MSG_EOR) call.
This patch handles the tcp_sendmsg case. The followup patches
will handle other skb coalescing and fragment cases.
One potential use case is to use MSG_EOR with
SOF_TIMESTAMPING_TX_ACK to get a more accurate
TCP ack timestamping on application protocol with
multiple outgoing response messages (e.g. HTTP2).
Signed-off-by: Martin KaFai Lau <kafai@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Soheil Hassas Yeganeh <soheil@google.com> Cc: Willem de Bruijn <willemb@google.com> Cc: Yuchung Cheng <ycheng@google.com> Suggested-by: Eric Dumazet <edumazet@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
tcp: remove SKBTX_ACK_TSTAMP since it is redundant
The SKBTX_ACK_TSTAMP flag is set in skb_shinfo->tx_flags when
the timestamp of the TCP acknowledgement should be reported on
error queue. Since accessing skb_shinfo is likely to incur a
cache-line miss at the time of receiving the ack, the
txstamp_ack bit was added in tcp_skb_cb, which is set iff
the SKBTX_ACK_TSTAMP flag is set for an skb. This makes
SKBTX_ACK_TSTAMP flag redundant.
Remove the SKBTX_ACK_TSTAMP and instead use the txstamp_ack bit
everywhere.
Note that this frees one bit in shinfo->tx_flags.
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com> Acked-by: Martin KaFai Lau <kafai@fb.com> Suggested-by: Willem de Bruijn <willemb@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
tcp: remove an unnecessary check in tcp_tx_timestamp
Remove the redundant check for sk->sk_tsflags in tcp_tx_timestamp.
tcp_tx_timestamp() receives the tsflags as a parameter. As a
result the "sk->sk_tsflags || tsflags" is redundant, since
tsflags already includes sk->sk_tsflags plus overrides from
control messages.
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Thu, 28 Apr 2016 13:33:24 +0000 (06:33 -0700)]
net: snmp: fix 64bit stats on 32bit arches
I accidentally replaced BH disabling by preemption disabling
in SNMP_ADD_STATS64() and SNMP_UPD_PO_STATS64() on 32bit builds.
For 64bit stats on 32bit arch, we really need to disable BH,
since the "struct u64_stats_sync syncp" might be manipulated
both from process and BH contexts.
Fixes: 6aef70a851ac ("net: snmp: kill various STATS_USER() helpers") Reported-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Tested-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Mon, 25 Apr 2016 17:39:32 +0000 (10:39 -0700)]
net: SOCKWQ_ASYNC_NOSPACE optimizations
SOCKWQ_ASYNC_NOSPACE is tested in sock_wake_async()
so that a SIGIO signal is sent when needed.
tcp_sendmsg() clears the bit.
tcp_poll() sets the bit when stream is not writeable.
We can avoid two atomic operations by first checking if socket
is actually interested in the FASYNC business (most sockets in
real applications do not use AIO, but select()/poll()/epoll())
This also removes one cache line miss to access sk->sk_wq->flags
in tcp_sendmsg()
Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
In the old days (before linux-3.0), SNMP counters were duplicated,
one set for user context, and anther one for BH context.
After commit 8f0ea0fe3a03 ("snmp: reduce percpu needs by 50%")
we have a single copy, and what really matters is preemption being
enabled or disabled, since we use this_cpu_inc() or __this_cpu_inc()
respectively.
This patch series kills the obsolete STATS_USER() helpers,
and rename all XXX_BH() helpers to __XXX() ones, to more
closely match conventions used to update per cpu variables.
This is probably going to hurt maintainers job for a while,
since cherry-picks will not be clean, but this had to be
cleaned at one point. I am so sorry guys.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
projects / karo-tx-linux.git / log