Toshiaki Makita says:
====================
Stacked vlan TSO
On the basis of Netdev 0.1 discussion[1], I made a patch set to enable
TSO for packets with multiple vlans.
Currently, packets with multiple vlans are always segmented by software,
which is caused by that netif_skb_features() drops most feature flags
for multiple tagged packets.
To allow NICs to segment them, we need to get rid of that check from core.
Fortunately, recently introduced ndo_features_check() can be used to
move the check to each driver, and this patch set is based on the idea.
For the initial patch set, I chose 3 drivers, bonding, team, and igb, as
candidates to enable TSO. I tested them and confirmed they works fine
with this change.
Here are samples of performance test results. As I expected, %sys gets
pretty lower than before.
* TEST1: vlan (.1Q) on vlan (.1ad) on igb (I350)
- before
$ netperf -t TCP_STREAM -H 192.168.10.1 -l 60
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 16384 16384 60.02 933.72
Average: CPU %user %nice %system %iowait %steal %idle
Average: all 0.13 0.00 11.28 0.01 0.00 88.58
- after
$ netperf -t TCP_STREAM -H 192.168.10.1 -l 60
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 16384 16384 60.01 936.13
Average: CPU %user %nice %system %iowait %steal %idle
Average: all 0.24 0.00 4.17 0.01 0.00 95.58
* TEST2: vlan (.1Q) on bridge (.1ad vlan filtering) on team on igb (I350)
- before
$ netperf -t TCP_STREAM -H 192.168.10.1 -l 60
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 16384 16384 60.01 936.28
Average: CPU %user %nice %system %iowait %steal %idle
Average: all 0.41 0.00 11.57 0.01 0.00 88.01
- after
$ netperf -t TCP_STREAM -H 192.168.10.1 -l 60
Recv Send Send
Socket Socket Message Elapsed
Size Size Size Time Throughput
bytes bytes bytes secs. 10^6bits/sec
87380 16384 16384 60.02 935.72
Average: CPU %user %nice %system %iowait %steal %idle
Average: all 0.14 0.00 7.66 0.01 0.00 92.19
In addition to above, I tested these configurations:
- vlan (.1Q) on vlan (1.ad) on bonding on igb (I350)
- vlan (.1Q) on vlan (1.Q) on igb (I350)
- vlan (.1Q) on vlan (1.Q) on team on igb (I350)
And didn't find any problem.
[1] https://netdev01.org/sessions/18https://netdev01.org/docs/netdev01_bof_8021ad_makita_150212.pdf
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
As datasheets for igb (I210, I350, 82576, etc.) say, maclen can be from
14 to 127, which is enough for reasonable number of vlan tags.
My netperf test showed I350's TSO works pretty fine with multiple vlans.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Team devices don't need to segment multiple tagged packets since their
slaves can segment them.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bonding devices don't need to segment multiple tagged packets since their
slaves can segment them.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
As there are a number of (especially virtual) devices that don't
need the multiple vlan check, introduce passthru_features_check() for
convenience.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
To allow drivers to handle the features check for multiple tags,
move the check to ndo_features_check().
As no drivers currently handle multiple tagged TSO, introduce
dflt_features_check() and call it if the driver does not have
ndo_features_check().
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Separate the two checks for single vlan and multiple vlans in
netif_skb_features(). This allows us to move the check for multiple
vlans to another function later.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stacked vlan devices curretly have few features (GRO, HIGHDMA, LLTX).
Since we have software fallbacks in case the NIC can not handle some
features for multiple vlans, we can add the same features as the lower
vlan devices for stacked vlan devices.
This allows stacked vlan devices to create large (GSO) packets and not to
segment packets. Those packets will be segmented by software on the real
device, or even can be segmented by the NIC once TSO for multiple vlans
becomes enabled by the following patches.
The exception is those related to FCoE, which does not have a software
fallback.
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes sparse warnings introduced in commit e85c9a7abf ("cxgb4/cxgb4vf: Add
code to calculate T5 BAR2 Offsets for SGE Queue Registers") and
df64e4d38c ("cxgb4/cxgb4vf: Use new interfaces to calculate BAR2 SGE Queue
Register addresses") and few old ones
sparse warnings:
>> drivers/net/ethernet/chelsio/cxgb4vf/sge.c:1006:48: sparse: cast removes
>> address space of expression
>> drivers/net/ethernet/chelsio/cxgb4vf/sge.c:1006:48: sparse: incorrect type in
>> initializer (different address space)
>> drivers/net/ethernet/chelsio/cxgb4vf/sge.c:1020:40: sparse: incorrect type in
>> argument 1 (different base types)
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
existing TC action 'pedit' can munge any bits of the packet.
Generalize it for use in bpf programs attached as cls_bpf and act_bpf via
bpf_skb_store_bytes() helper function.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Guenter Roeck says:
====================
net: dsa: HW bridging, EEE support
Patch 1 to 7 of this series prepare the drivers using the mv88e6xxx code
for HW bridging support, without adding the code itself. For the most part
this factors out common port initialization code. There is no functional
change except for patch 3, which disables the message port bit for the
CPU port to prevent packet duplication if HW bridging is configured.
Patch 8 adds the infrastructure for hardware bridging support to the
mv88e6xxx code.
Patch 9 wires the MV88E6352 driver to support hardware bridging.
Patches 10 to 12 add support for ndo_fdb functions to the dsa subsystem,
and wire up the MV88E6352 driver to support those functions.
Patches 13 to 16 add EEE support and HW bridging support to the mv88e6171
driver. This set of patches is from Andrew, applied on top of the first
set of patches.
The series applies to net-next as of 3/24/2015.
Thanks a lot to Andrew Lunn for testing and valuable feedback.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Wire up the common code for setting up hardware bridging
and access to the forwarding database.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mv88e6172 has support for EEE. Check for the product ID and call
the common code if applicable.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make the code more readable by using defines for the switch IDs.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get the switch id and save it away in the private mv88x6xxx structure
in a centralised piece of code, rather than each driver doing it itself.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for manipulating switch fdb entries by pointing to the
ndo_fdb functions implemented for mv88e6xxxx.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
No vlan support at this time.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provide callbacks for ndo_fdb_add, ndo_fdb_del, and ndo_fdb_dump.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bridge support is similar for all chips supported by the mv88e6xxx code,
so add the code there.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This will simplify adding offloaded bridge support later on.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This prepares the driver for hardware bridging.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split mv88e6xxx_reg_read and mv88e6xxx_reg_write into two functions each,
one to acquire smi_mutex and one to get struct mii_bus *bus from
struct dsa_switch *ds and to call the actual read/write function.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Datasheet says that the Message Port bit should not be set for the CPU port.
Having it set causes DSA tagged packets to be sent to the CPU port roughly
every 30 seconds. Those packets are the same as real packets forwarded between
switch ports if the switch is configured for switching between multiple ports.
The packets are then bridged by the software bridge, resulting in duplicated
packets on the network.
Reported-by: Andrew Lunn <andrew@lunn.ch>
Cc: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provide mv88e6xxx_setup_port_common() for common port initialization.
Currently only write Port 1 Control and VLAN configuration since
this will be needed for hardware bridging. More can be added later
if desired/needed.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Code used and needed in mv886xxx.c should be initialized there as well,
so factor it out from the individual initialization files.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vmbus_are_subchannels_present() also involves opening the channels, which
may be too early at this point. Checking for subchannels is not necessary here.
So this patch removes it. Subchannels will be opened when offer messages arrive.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since 39b2bbe3d7 (gpio: add flags argument to gpiod_get*() functions)
which appeared in v3.17-rc1, the gpiod_get* functions take an additional
parameter that allows to specify direction and initial value for output.
Simplify accordingly.
Moreover use devm_gpiod_get_index_optional for still simpler handling.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the current code, ids are removed too early.
Suppose you have an ipip interface that stands in the netns foo and its link
part in the netns bar (so the netns bar has an nsid into the netns foo).
Now, you remove the netns bar:
- the bar nsid into the netns foo is removed
- the netns exit method of ipip is called, thus our ipip iface is removed:
=> a netlink message is sent in the netns foo to advertise this deletion
=> this netlink message requests an nsid for bar, thus a new nsid is
allocated for bar and never removed.
We must remove nsids when we are sure that nobody will refer to netns currently
cleaned.
Fixes: 0c7aecd4bd ("netns: add rtnl cmd to add and get peer netns ids")
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With this patch, we can send out multiple RNDIS data packets in one send buffer
slot and one VMBus message. It reduces the overhead associated with VMBus messages.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains Netfilter updates for your net-next tree.
Basically, nf_tables updates to add the set extension infrastructure and finish
the transaction for sets from Patrick McHardy. More specifically, they are:
1) Move netns to basechain and use recently added possible_net_t, from
Patrick McHardy.
2) Use LOGLEVEL_<FOO> from nf_log infrastructure, from Joe Perches.
3) Restore nf_log_trace that was accidentally removed during conflict
resolution.
4) nft_queue does not depend on NETFILTER_XTABLES, starting from here
all patches from Patrick McHardy.
5) Use raw_smp_processor_id() in nft_meta.
Then, several patches to prepare ground for the new set extension
infrastructure:
6) Pass object length to the hash callback in rhashtable as needed by
the new set extension infrastructure.
7) Cleanup patch to restore struct nft_hash as wrapper for struct
rhashtable
8) Another small source code readability cleanup for nft_hash.
9) Convert nft_hash to rhashtable callbacks.
And finally...
10) Add the new set extension infrastructure.
11) Convert the nft_hash and nft_rbtree sets to use it.
12) Batch set element release to avoid several RCU grace period in a row
and add new function nft_set_elem_destroy() to consolidate set element
release.
13) Return the set extension data area from nft_lookup.
14) Refactor existing transaction code to add some helper functions
and document it.
15) Complete the set transaction support, using similar approach to what we
already use, to activate/deactivate elements in an atomic fashion.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Ying Xue says:
====================
tipc: fix two corner issues
The patch set aims at resolving the following two critical issues:
Patch #1: Resolve a deadlock which happens while all links are reset
Patch #2: Correct a mistake usage of RCU lock which is used to protect
node list
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
TIPC node hash node table is protected with rcu lock on read side.
tipc_node_find() is used to look for a node object with node address
through iterating the hash node table. As the entire process of what
tipc_node_find() traverses the table is guarded with rcu read lock,
it's safe for us. However, when callers use the node object returned
by tipc_node_find(), there is no rcu read lock applied. Therefore,
this is absolutely unsafe for callers of tipc_node_find().
Now we introduce a reference counter for node structure. Before
tipc_node_find() returns node object to its caller, it first increases
the reference counter. Accordingly, after its caller used it up,
it decreases the counter again. This can prevent a node being used by
one thread from being freed by another thread.
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericson.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
received is 0, no need to minus it and use "+=" to reassign it
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sathya Perla says:
====================
be2net: patch set
Hi David, this patch set includes 2 feature additions to the be2net driver:
Patch 1 sets up cpu affinity hints for be2net irqs using the
cpumask_set_cpu_local_first() API that first picks the near numa cores
and when they are exhausted, selects the far numa cores.
Patch 2 setups up xps queue mapping for be2net's TXQs to avoid,
by default, TX lock contention.
Patch 3 just bumps up the driver version.
Pls consider applying this patch set to the net-next queue. Thanks!
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch sets up xps queue mapping on load, so that TX traffic is
steered to the queue whose irqs are being processed by the current cpu.
This helps in avoiding TX lock contention.
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch provides hints to irqbalance to map be2net IRQs to
specific CPU cores. cpumask_set_cpu_local_first() is used, which first
maps IRQs to near NUMA cores; when those cores are exhausted, IRQs are
mapped to far NUMA cores.
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@emulex.com>
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the ITS or the redistributors report their shareability as zero,
then it is important to make sure they will no generate any cacheable
traffic, as this is unlikely to produce the expected result.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Link: https://lkml.kernel.org/r/1427465705-17126-5-git-send-email-marc.zyngier@arm.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The ITS driver sometime mixes up the use of GICR_PROPBASE bitfields
for the GICR_PENDBASE register, and GITS_BASER for GICR_CBASE.
This does not lead to any observable bug because similar bits are
at the same location, but this just make the code even harder to
understand...
This patch provides the required #defines and fixes the mixup.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Link: https://lkml.kernel.org/r/1427465705-17126-4-git-send-email-marc.zyngier@arm.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
When building ITS commands which have the device ID in it, we
should mask off the whole upper 32 bits of the first command word
before inserting the new value in there.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Link: https://lkml.kernel.org/r/1427465705-17126-3-git-send-email-marc.zyngier@arm.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
With a monolithic GICv3, redistributors are addressed using a linear
number, while a distributed implementation uses physical addresses.
When encoding a target address into a command, we strip the lower
16 bits, as redistributors are always 64kB aligned. This works
perfectly well with a distributed implementation, but has the
silly effect of always encoding target 0 in the monolithic case
(unless you have more than 64k CPUs, of course).
The obvious fix is to shift the linear target number by 16 when
computing the target address, so that we don't loose any precious
bit.
Reported-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Link: https://lkml.kernel.org/r/1427465705-17126-2-git-send-email-marc.zyngier@arm.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Hariprasad Shenai says:
====================
cxgb4: Fixes ingress queue mapping and other fixes
The below series fixes ingress queue mapping by allocating them dynamically to
prevent stack overflow. Disable napi and interrupts before unregistering netdev
to avoid crash while unloading driver when traffic is flowing.
The patches series is created against 'net' tree.
And includes patches on cxgb4 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>
Disable interrupts and quiesce rx before unregistering net device to avoid crash
while unloading driver when traffic is flowing through.
Based on original work by Shameem Khalid <shameem@chelsio.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
QIDs (egress/ingress) from firmware in FW_*_CMD.alloc command
can be anywhere in the range from EQ(IQFLINT)_START to EQ(IQFLINT)_END.
For eg, in the first load eqid can be from 100 to 300.
In the next load it can be from 301 to 500 (assume eq_start is 100 and eq_end is
1000).
The driver was assuming them to always start from EQ(IQFLINT)_START till
MAX_EGRQ(INGQ). This was causing stack overflow and subsequent crash.
Fixed it by dynamically allocating memory (of qsize (x_END - x_START + 1)) for
these structures.
Based on original work by Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 1fb6f159fd ("tcp: add tcp_conn_request"),
tcp_syn_flood_action() is no longer used from IPv6.
We can make it static, by moving it above tcp_conn_request()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Octavian Purdila <octavian.purdila@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/chelsio/cxgb4/cxgb4_fcoe.c:49:9-10: WARNING: return of 0/1 in function 'cxgb_fcoe_sof_eof_supported' with return type bool
Return statements in functions returning bool should use
true/false instead of 1/0.
Generated by: scripts/coccinelle/misc/boolreturn.cocci
CC: Varun Prakash <varun@chelsio.com>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller