This patch causes the proper attribute flags to be set,
in the case that IPv6 UDP checksums are disabled, so that
userspace ie. `ip l2tp show tunnel` knows about it.
Signed-off-by: Asbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only set L2TP_ATTR_UDP_CSUM in l2tp_nl_tunnel_send()
when it's running over IPv4.
This prepares the code to also have IPv6 specific attributes.
Signed-off-by: Asbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: David S. Miller <davem@davemloft.net>
The attributes L2TP_ATTR_UDP_ZERO_CSUM6_RX and
L2TP_ATTR_UDP_ZERO_CSUM6_TX are used as flags,
but is defined as a u8 in a comment.
This patch redocuments them as flags.
Adding nla_policy entries would break API, so not doing that.
CC: Tom Herbert <therbert@google.com>
Signed-off-by: Asbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: David S. Miller <davem@davemloft.net>
Receiving a GSO packet in dev_gro_receive() is not uncommon
in stacked devices, or devices partially implementing LRO/GRO
like bnx2x. GRO is implementing the aggregation the device
was not able to do itself.
Current code causes reorders, like in following case :
For a given flow where sender sent 3 packets P1,P2,P3,P4
Receiver might receive P1 as a single packet, stored in GRO engine.
Then P2-P4 are received as a single GSO packet, immediately given to
upper stack, while P1 is held in GRO engine.
This patch will make sure P1 is given to upper stack, then P2-P4
immediately after.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Edward Cree says:
====================
sfc: enable 4-tuple UDP RSS hashing
EF10 based NICs have configurable RSS hash fields, and can be made to take the
ports into the hash on UDP (they already do so for TCP). This patch series
enables this, in order to improve spreading of UDP traffic.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This improves UDP spreading, and also slightly improves GRO performance
of encapsulated TCP on 7000 series NICs.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
Mellanox 100G SRIOV offloads tunnel_key set/release
From Hadar Hen Zion:
This series further enhances the SRIOV TC offloads of mlx5 to handle the
TC tunnel_key release and set actions.
This serves a common use-case in virtualization systems where the virtual
switch encapsulate packets (tunnel_key set action) sent from VMs with
outer headers corresponding to the local/remote host IPs and de-capsulate
(tunnel_key release) outer headers before the packets are received by the
VM.
We use the new E-Switch switchdev mode and TC tunnel_key set/release
action to achieve that also in SW defined SRIOV environments by
offloading TC rules that contain these actions along with forwarding
(TC mirred/redirect action) the packets.
The first six patches are adding the needed support in flow dissector,
flower and tc for offloading tunnel_key actions:
- The first three patches are adding the needed help functions
and enums
- The next three patches in the series are adding UDP port attribute
to tunnel_key release and set actions.
The addition of UDP ports would allow the HW driver to make sure they are
given (say) a VXLAN tunnel to offload (mlx5e uses that).
Patches 7-10 are mlx5 preparations for tunnel_key actions offloads support.
Patch #11 adds mlx5e support to offload tunnel_key release action, and the
last two patches (#12-13) add mlx5e support to tc tunnel_key set action.
Currently in order to offload tc tunnel_key release action, the tc rule
should be placed on top of the mlx5e offloading (uplink) interface instead
of the shared tunnel interface. The resolution between the tunnel interface
to the HW netdevice will be implemented in a follow up series.
This series was generated against commit
94edc86bf1 ("Merge branch 'dwmac-sti-refactor-cleanup'")
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
In mlx5 HW, encapsulation is offloaded by the steering rule having
index into an encapsulation table containing the entire set of headers
to be added by the HW. The driver sets these headers in a buffer when we
are offloading the action.
The code maintains mlx5_encap_entry for each encap header it has
encountered when attempted to offload TC tunnel set action.
This entry maintains a linked list of all the flows sharing the same
encap header, when the last flow is removed from the list the encap
entry is removed.
The actual encap_header is allocated by the driver in the hardware only
if we have layer two neighbour info when the encap entry is created.
While the flow is in the driver, the driver holds a reference on the
neighbour.
When a new flow with encap action is inserted, the code first checks if
the required encap entry exists according to the tunnel set parameters.
If it does the encap is shared, otherwise a new mlx5_encap_entry is
created.
TC action parsing implementation in the driver assumes that tunnel set
action is provided in the same order set by the user, e.g before the
mirred_redirect action.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By implementing this ndo, the host stack will set the vxlan udp port
also to VF representor netdevices. This will allow the TC offload code
in the driver when it gets a tunnel key set action to identify the UDP
port as vxlan, and hence the rule will be a candidate for offloading.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Enhance the parsing of offloaded TC rules to set HW matching on outer
(encapsulation) headers.
Parse TC tunnel release action and set it as mlx5 decap action when the
required capabilities are supported.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to support steering rules which add encapsulation headers,
encap_id parameter is needed.
Add new mlx5_flow_act struct which holds action related parameter:
action, flow_tag and encap_id. Use mlx5_flow_act struct when adding a new
steering rule.
This patch doesn't change any functionality.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When creating flow tables, allow the caller to specify creation flags.
Currently no flags are used and as such this patch doesn't add any new
functionality.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of comparing to a const value, check the value of max encap
header size capability as reported by the Firmware.
Fixes: 575ddf5888 ('net/mlx5: Introduce alloc_encap and dealloc_encap commands')
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The alloc and dealloc encap commands will be used in the mlx5e driver,
as such, declare them in a common header file.
Also, rename the functions: mlx5_cmd_{de}alloc_encap is replaced with
mlx5_encap_{de}alloc.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current tunnel set action supports only IP addresses and key
options. Add UDP dst port option.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add dst port parameter to __ip_tun_set_dst and __ipv6_tun_set_dst
utility functions.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current IP tunneling classification supports only IP addresses and key.
Enhance UDP based IP tunneling classification parameters by adding UDP
src and dst port.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When encapsulation field is set, mark it as used key for the flow
dissector. This will be used by offloading drivers.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
New encapsulation keys were added to the flower classifier, which allow
classification according to outer (encapsulation) headers attributes
such as key and IP addresses.
In order to expose those attributes outside flower, add
corresponding enums in the flow dissector.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Needed for drivers to pick the relevant action when offloading tunnel
key act.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Without this check, it is not possible to create two rules that
are identical except for their UID ranges. For example:
root@net-test:/# ip rule add prio 1000 lookup 300
root@net-test:/# ip rule add prio 1000 uidrange 100-200 lookup 300
RTNETLINK answers: File exists
root@net-test:/# ip rule add prio 1000 uidrange 100-199 lookup 100
root@net-test:/# ip rule add prio 1000 uidrange 200-299 lookup 200
root@net-test:/# ip rule add prio 1000 uidrange 300-399 lookup 100
RTNETLINK answers: File exists
Tested: https://android-review.googlesource.com/#/c/299980/
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Acked-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Driver uses a union for copying data to & from management firmware
when interacting with it.
Problem is that the function always copies sizeof(union) while commit
2edbff8dcb ("qed: Learn resources from management firmware") is casting
a union elements which is of smaller size [24-byte instead of 88-bytes].
Also, the union contains some inappropriate elements which increase its
size [should have been 32-bytes]. While this shouldn't corrupt other
PF messages to the MFW [as management firmware enforces permissions so
that each PF is allowed to write only to its own mailbox] we fix this
here as well.
Fixes: 2edbff8dcb ("qed: Learn resources from management firmware")
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When moving from typhoon_get_settings to typhoon_getlink_ksettings
in the commit f7a5537cd2 ("net: 3com: typhoon: use new api
ethtool_{get|set}_link_ksettings"), we use a local variable supported
but we forgot to update the struct ethtool_link_ksettings with
this value.
We also initialize advertising to zero, because otherwise it may
be uninitialized if no case of the switch (tp->xcvr_select) is used.
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>
Acked-by: Tom Lendacky <thomas.lendacky@amd.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>
Acked-by: Jes Sorensen <Jes.Sorensen@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>
Joachim Eastwood says:
====================
stmmac: dwmac-rk: convert to standard PM/remove functions
This patch set aims to remove the init/exit callbacks from the
dwmac-rk driver and instead use standard PM callbacks. Eventually
the init/exit callbacks will be deprecated and removed from all
drivers dwmac-* except for dwmac-generic. Drivers will be refactored
to use standard PM and remove callbacks.
This conversion was pretty straight forward, but it would really nice
if some chromium people could test suspend/resume with this patch set.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of adding hooks inside stmmac_platform it is better to just use
the standard PM callbacks within the specific dwmac-driver. This only
used by the dwmac-rk driver.
This reverts commit cecbc5563a ("stmmac: allow to split suspend/resume
from init/exit callbacks").
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since the rk_gmac_init() only calls another function move this
function call into probe so rk_gmac_init() can be removed.
Since commit cecbc5563a ("stmmac: allow to split suspend/resume
from init/exit callbacks") the init hook is no longer used in
dwmac-rk so this can be removed.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert the exit hook into a standard driver remove function as
the hook doesn't really buy us anything extra.
Eventually the exit hook will be deprecated in favor of the driver
remove function.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use standard PM resume/suspend callbacks instead of the hooks in
stmmac_platform. This gives the driver more control and flexibility
when implementing PM functionality. The hooks in stmmac_platform
also doesn't buy us anything extra.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp: tcp_get_info() locking changes
This short series prepares tcp_get_info() for more detailed infos.
In order to not slow down fast path, our goal is to use the normal
socket spinlock instead of custom synchronization.
All we need to ensure is that tcp_get_info() is not called with
ehash lock, which might dead lock, since packet processing would acquire
the spinlocks in reverse way.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
We had various problems in the past in tcp_get_info() and used
specific synchronization to avoid deadlocks.
We would like to add more instrumentation points for TCP, and
avoiding grabing socket lock in tcp_getinfo() was too costly.
Being able to lock the socket allows to provide consistent set
of fields.
inet_diag_dump_icsk() can make sure ehash locks are not
held any more when tcp_get_info() is called.
We can remove syncp added in commit d654976cbf
("tcp: fix a potential deadlock in tcp_get_info()"), but we need
to use lock_sock_fast() instead of spin_lock_bh() since TCP input
path can now be run from process context.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Being lockless in tcp_get_info() is hard, because we need to add
specific synchronization in TCP fast path, like seqcount.
Following patch will change inet_diag_dump_icsk() to no longer
hold any lock for non listeners, so that we can properly acquire
socket lock in get_tcp_info() and let it return more consistent counters.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Neil Armstrong says:
====================
ARM64: Add Internal PHY support for Meson GXL
The Amlogic Meson GXL SoCs have an internal RMII PHY that is muxed with the
external RGMII pins.
In order to support switching between the two PHYs links, extended registers
size for mdio-mux-mmioreg must be added.
The DT related patches submitted as RFC in [3] will be sent in a separate
patchset due to multiple patchsets and DTSI migrations.
Changes since v2 RFC patchset at : [3]
- Change phy Kconfig/Makefile alphabetic order
- GXL dtsi cleanup
Changes since original RFC patchset at : [2]
- Remove meson8b experimental phy switching
- Switch to mdio-mux-mmioreg with extennded size support
- Add internal phy support for S905x and p231
- Add external PHY support for p230
[1] http://lkml.kernel.org/r/1477932286-27482-1-git-send-email-narmstrong@baylibre.com
[2] http://lkml.kernel.org/r/1477060838-14164-1-git-send-email-narmstrong@baylibre.com
[3] http://lkml.kernel.org/r/1477932987-27871-1-git-send-email-narmstrong@baylibre.com
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add driver for the Internal RMII PHY found in the Amlogic Meson GXL SoCs.
This PHY seems to only implement some standard registers and need some
workarounds to provide autoneg values from vendor registers.
Some magic values are currently used to configure the PHY, and this a
temporary setup until clarification about these registers names and
registers fields are provided by Amlogic.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to support PHY switching on Amlogic GXL SoCs, add support for
16bit and 32bit registers sizes.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sowmini Varadhan says:
====================
RDS: TCP: bug fixes
A couple of bug fixes identified during testing.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The for() loop in rds_tcp_accept_one() assumes that the 0'th
rds_tcp_conn_path is UP and starts multipath accepts at index 1.
But this assumption may not always be true: if the 0'th path
has failed (ERROR or DOWN state) an incoming connection request
should be used to resurrect this path.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The socket argument passed to rds_tcp_tc_info() is a PF_RDS socket,
so it is incorrect to report the address port info based on
rds_getname() as part of TCP state report.
Invoke inet_getname() for the t_sock associated with the
rds_tcp_connection instead.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Do not set sk_err when dequeuing errors from the error queue.
Doing so results in:
a) Bugs: By overwriting existing sk_err values, it possibly
hides legitimate errors. It is also incorrect when local
errors are queued with ip_local_error. That happens in the
context of a system call, which already returns the error
code.
b) Inconsistent behavior: When there are pending errors on
the error queue, sk_err is sometimes 0 (e.g., for
the first timestamp on the error queue) and sometimes
set to an error code (after dequeuing the first
timestamp).
c) Suboptimality: Setting sk_err to ENOMSG on simple
TX timestamps can abort parallel reads and writes.
Removing this line doesn't break userspace. This is because
userspace code cannot rely on sk_err for detecting whether
there is something on the error queue. Except for ICMP messages
received for UDP and RAW, sk_err is not set at enqueue time,
and as a result sk_err can be 0 while there are plenty of
errors on the error queue.
For ICMP packets in UDP and RAW, sk_err is set when they are
enqueued on the error queue, but that does not result in aborting
reads and writes. For such cases, sk_err is only readable via
getsockopt(SO_ERROR) which will reset the value of sk_err on
its own. More importantly, prior to this patch,
recvmsg(MSG_ERRQUEUE) has a race on setting sk_err (i.e.,
sk_err is set by sock_dequeue_err_skb without atomic ops or
locks) which can store 0 in sk_err even when we have ICMP
messages pending. Removing this line from sock_dequeue_err_skb
eliminates that race.
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jesper Dangaard Brouer says:
====================
qdisc and tx_queue_len cleanups for IFF_NO_QUEUE devices
This patchset is a cleanup for IFF_NO_QUEUE devices. It will
hopefully help userspace get a more consistent behavior when attaching
qdisc to such virtual devices.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
It is a clear misconfiguration to attach a qdisc to a device with
tx_queue_len zero, because some qdisc's (namely, pfifo, bfifo, gred,
htb, plug and sfb) inherit/copy this value as their queue length.
Why should the kernel catch such a misconfiguration? Because prior to
introducing the IFF_NO_QUEUE device flag, userspace found a loophole
in the qdisc config system that allowed them to achieve the equivalent
of IFF_NO_QUEUE, which is to remove the qdisc code path entirely from
a device. The loophole on older kernels is setting tx_queue_len=0,
*prior* to device qdisc init (the config time is significant, simply
setting tx_queue_len=0 doesn't trigger the loophole).
This loophole is currently used by Docker[1] to get better performance
and scalability out of the veth device. The Docker developers were
warned[1] that they needed to adjust the tx_queue_len if ever
attaching a qdisc. The OpenShift project didn't remember this warning
and attached a qdisc, this were caught and fixed in[2].
[1] https://github.com/docker/libcontainer/pull/193
[2] https://github.com/openshift/origin/pull/11126
Instead of fixing every userspace program that used this loophole, and
forgot to reset the tx_queue_len, prior to attaching a qdisc. Let's
catch the misconfiguration on the kernel side.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The flag IFF_NO_QUEUE marks virtual device drivers that doesn't need a
default qdisc attached, given they will be backed by physical device,
that already have a qdisc attached for pushback.
It is still supported to attach a qdisc to a IFF_NO_QUEUE device, as
this can be useful for difference policy reasons (e.g. bandwidth
limiting containers). For this to work, the tx_queue_len need to have
a sane value, because some qdiscs inherit/copy the tx_queue_len
(namely, pfifo, bfifo, gred, htb, plug and sfb).
Commit a813104d92 ("IFF_NO_QUEUE: Fix for drivers not calling
ether_setup()") caught situations where some drivers didn't initialize
tx_queue_len. The problem with the commit was choosing 1 as the
fallback value.
A qdisc queue length of 1 causes more harm than good, because it
creates hard to debug situations for userspace. It gives userspace a
false sense of a working config after attaching a qdisc. As low
volume traffic (that doesn't activate the qdisc policy) works,
like ping, while traffic that e.g. needs shaping cannot reach the
configured policy levels, given the queue length is too small.
This patch change the value to DEFAULT_TX_QUEUE_LEN, given other
IFF_NO_QUEUE devices (that call ether_setup()) also use this value.
Fixes: a813104d92 ("IFF_NO_QUEUE: Fix for drivers not calling ether_setup()")
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The default TX queue length of Ethernet devices have been a magic
constant of 1000, ever since the initial git import.
Looking back in historical trees[1][2] the value used to be 100,
with the same comment "Ethernet wants good queues". The commit[3]
that changed this from 100 to 1000 didn't describe why, but from
conversations with Robert Olsson it seems that it was changed
when Ethernet devices went from 100Mbit/s to 1Gbit/s, because the
link speed increased x10 the queue size were also adjusted. This
value later caused much heartache for the bufferbloat community.
This patch merely moves the value into a defined constant.
[1] https://git.kernel.org/cgit/linux/kernel/git/davem/netdev-vger-cvs.git/
[2] https://git.kernel.org/cgit/linux/kernel/git/tglx/history.git/
[3] https://git.kernel.org/tglx/history/c/98921832c232
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Paolo Abeni says:
====================
udp: do fwd memory scheduling on dequeue
After commit 850cbaddb5 ("udp: use it's own memory accounting schema"),
the udp code needs to acquire twice the receive queue spinlock on dequeue.
This patch series remove the need for the second lock at skb free time,
moving the udp memory scheduling inside the dequeue operation; the skb
destructor field is not used anymore and an additional sk argument is added
to ip_cmsg_recv_offset() to cope with null skb->sk after dequeue.
Many thanks to Eric Dumazed for suggesting pretty all much the above.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
A new argument is added to __skb_recv_datagram to provide
an explicit skb destructor, invoked under the receive queue
lock.
The UDP protocol uses such argument to perform memory
reclaiming on dequeue, so that the UDP protocol does not
set anymore skb->desctructor.
Instead explicit memory reclaiming is performed at close() time and
when skbs are removed from the receive queue.
The in kernel UDP protocol users now need to call a
skb_recv_udp() variant instead of skb_recv_datagram() to
properly perform memory accounting on dequeue.
Overall, this allows acquiring only once the receive queue
lock on dequeue.
Tested using pktgen with random src port, 64 bytes packet,
wire-speed on a 10G link as sender and udp_sink as the receiver,
using an l4 tuple rxhash to stress the contention, and one or more
udp_sink instances with reuseport.
nr sinks vanilla patched
1 440 560
3 2150 2300
6 3650 3800
9 4450 4600
12 6250 6450
v1 -> v2:
- do rmem and allocated memory scheduling under the receive lock
- do bulk scheduling in first_packet_length() and in udp_destruct_sock()
- avoid the typdef for the dequeue callback
Suggested-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Asbjørn Sloth Tønnesen