XDP_TX requires a different function to handle completion. Add a
function pointer to handle tx completion logic. Regular TX rings
will be assigned the current bnxt_tx_int() for the ->tx_int()
function pointer.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add logic for an extra set of TX rings for XDP. If enabled, this
set of TX rings equals the number of RX rings and shares the same
IRQ as the RX ring set. A new field bp->tx_nr_rings_xdp is added
to keep track of these TX XDP rings. Adjust all other relevant functions
to handle bp->tx_nr_rings_xdp.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To support XDP_TX, we need to add a set of dedicated TX rings, each
associated with the NAPI of an RX ring. To assign XDP rings and regular
rings in a flexible way, we add a bp->tx_ring_map[] array to do the
remapping. The netdev txq index is stored in the new field txq_index
so that we can retrieve the netdev txq when handling TX completions.
In this patch, before we introduce XDP_TX, the mapping is 1:1.
v2: Fixed a bug in bnxt_tx_int().
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, bnxt_setup_tc() and bnxt_set_channels() have similar and
duplicated code to check and reserve rx and tx rings. Add a new
function bnxt_reserve_rings() to centralize the logic. This will
make it easier to add XDP_TX support which requires allocating a
new set of TX rings.
Also, the tx ring checking logic in bnxt_setup_msix() can be removed.
The rings have been reserved before hand.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the current code, we have separate rx_event and agg_event parameters
to keep track of rx and aggregation events. Combine these events into
an u8 event mask with different bits defined for different events. This
way, it is easier to expand the logic to include XDP tx events.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This mode is to support XDP. In this mode, each rx ring is configured
with page sized buffers for linear placement of each packet. MTU will be
restricted to what the page sized buffers can support.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert the global constants BNXT_RX_OFFSET and BNXT_RX_DMA_OFFSET to
device parameters. This will make it easier to support XDP with
headroom support which requires different RX buffer offsets.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When driver is running in XDP mode, rx buffers are DMA mapped as
DMA_BIDIRECTIONAL. Add a field so the code will map/unmap rx buffers
according to this field.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To support XDP_TX, we need the RX buffer's DMA address to transmit the
packet. Convert the DMA address field to a permanent field in
bnxt_sw_rx_bd.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor refactoring of bnxt_rx_skb() so that it can easily be replaced by
a new function that handles packets in a single page. Also, use a
function pointer bp->rx_skb_func() to switch to a new function when
we add the new mode in the next patch.
Add a new field data_ptr that points to the packet data in the
bnxt_sw_rx_bd structure. The original data field is changed to void
pointer so that it can either hold the kmalloc'ed data or a page
pointer.
The last parameter of bnxt_rx_skb() which was the length parameter is
changed to include the payload offset of the packet in the upper 16 bit.
The offset is needed to support the rx page mode and is not used in
this existing function.
v3: Added a new data_ptr parameter to bp->rx_skb_func(). The caller
has the option to modify the starting address of the packet. This
will be needed when XDP with headroom support is added.
v2: Changed the name of the last parameter to offset_and_len to make the
code more clear.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To support setting the pause parameters, the driver can no longer just
mirror the PHY. The set_pauseparam feature allows the driver to
force the setting in the MAC, regardless of how the PHY is configured.
This means that we now need to maintain an internal state for pause
frame support, and so get_pauseparam also needs to be updated.
If the interface is already running when the setting is changed, then
the interface is reset.
Note that if the MAC is configured to enable RX pause frame support
(i.e. it transmits pause frames to throttle the other end), but the
PHY is configured to block those frames, then the feature will not work.
Also some buffer size initialization code into emac_init_adapter(),
so that it lives with similar code, including the initializtion of
pause frame support.
Signed-off-by: Timur Tabi <timur@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Julian Anastasov says:
====================
net: dst_confirm replacement
This patchset addresses the problem of neighbour
confirmation where received replies from one nexthop
can cause confirmation of different nexthop when using
the same dst. Thanks to YueHaibing <yuehaibing@huawei.com>
for tracking the dst->pending_confirm problem.
Sockets can obtain cached output route. Such
routes can be to known nexthop (rt_gateway=IP) or to be
used simultaneously for different nexthop IPs by different
subnet prefixes (nh->nh_scope = RT_SCOPE_HOST, rt_gateway=0).
At first look, there are more problems:
- dst_confirm() sets flag on dst and not on dst->path,
as result, indication is lost when XFRM is used
- DNAT can change the nexthop, so the really used nexthop is
not confirmed
So, the following solution is to avoid using
dst->pending_confirm.
The current dst_confirm() usage is as follows:
Protocols confirming dst on received packets:
- TCP (1 dst per socket)
- SCTP (1 dst per transport)
- CXGB*
Protocols supporting sendmsg with MSG_CONFIRM [ | MSG_PROBE ] to
confirm neighbour:
- UDP IPv4/IPv6
- ICMPv4 PING
- RAW IPv4/IPv6
- L2TP/IPv6
MSG_CONFIRM for other purposes (fix not needed):
- CAN
Sending without locking the socket:
- UDP (when no cork)
- RAW (when hdrincl=1)
Redirects from old to new GW:
- rt6_do_redirect
The patchset includes the following changes:
1. sock: add sk_dst_pending_confirm flag
- used only by TCP with patch 4 to remember the received
indication in sk->sk_dst_pending_confirm
2. net: add dst_pending_confirm flag to skbuff
- skb->dst_pending_confirm will be used by all protocols
in following patches, via skb_{set,get}_dst_pending_confirm
3. sctp: add dst_pending_confirm flag
- SCTP uses per-transport dsts and can not use
sk->sk_dst_pending_confirm like TCP
4. tcp: replace dst_confirm with sk_dst_confirm
5. net: add confirm_neigh method to dst_ops
- IPv4 and IPv6 provision for slow neigh lookups for MSG_PROBE users.
I decided to use neigh lookup only for this case because on
MSG_PROBE the skb may pass MTU checks but it does not reach
the neigh confirmation code. This patch will be used from patch 6.
- xfrm_confirm_neigh: we use the last tunnel address, if present.
When there are only transports, the original dest address is used.
6. net: use dst_confirm_neigh for UDP, RAW, ICMP, L2TP
- dst_confirm conversion for UDP, RAW, ICMP and L2TP/IPv6
- these protocols use MSG_CONFIRM propagated by ip*_append_data
to skb->dst_pending_confirm. sk->sk_dst_pending_confirm is not
used because some sending paths do not lock the socket. For
MSG_PROBE we use the slow lookup (dst_confirm_neigh).
- there are also 2 cases that need the slow lookup:
__ip6_rt_update_pmtu and rt6_do_redirect. I hope
&ipv6_hdr(skb)->saddr is the correct nexthop address to use here.
7. net: pending_confirm is not used anymore
- I failed to understand the CXGB* code, I see dst_confirm()
calls but I'm not sure dst_neigh_output() was called. For now
I just removed the dst->pending_confirm flag and left all
dst_confirm() calls there. Any better idea?
- Now may be old function neigh_output() should be restored
instead of dst_neigh_output?
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
As last step, we can remove the pending_confirm flag.
Reported-by: YueHaibing <yuehaibing@huawei.com>
Fixes: 5110effee8 ("net: Do delayed neigh confirmation.")
Fixes: f2bb4bedf3 ("ipv4: Cache output routes in fib_info nexthops.")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
The datagram protocols can use MSG_CONFIRM to confirm the
neighbour. When used with MSG_PROBE we do not reach the
code where neighbour is confirmed, so we have to do the
same slow lookup by using the dst_confirm_neigh() helper.
When MSG_PROBE is not used, ip_append_data/ip6_append_data
will set the skb flag dst_pending_confirm.
Reported-by: YueHaibing <yuehaibing@huawei.com>
Fixes: 5110effee8 ("net: Do delayed neigh confirmation.")
Fixes: f2bb4bedf3 ("ipv4: Cache output routes in fib_info nexthops.")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add confirm_neigh method to dst_ops and use it from IPv4 and IPv6
to lookup and confirm the neighbour. Its usage via the new helper
dst_confirm_neigh() should be restricted to MSG_PROBE users for
performance reasons.
For XFRM prefer the last tunnel address, if present. With help
from Steffen Klassert.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
Use the new sk_dst_confirm() helper to propagate the
indication from received packets to sock_confirm_neigh().
Reported-by: YueHaibing <yuehaibing@huawei.com>
Fixes: 5110effee8 ("net: Do delayed neigh confirmation.")
Fixes: f2bb4bedf3 ("ipv4: Cache output routes in fib_info nexthops.")
Tested-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new transport flag to allow sockets to confirm neighbour.
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
The flag is propagated from transport to every packet.
It is reset when cached dst is reset.
Reported-by: YueHaibing <yuehaibing@huawei.com>
Fixes: 5110effee8 ("net: Do delayed neigh confirmation.")
Fixes: f2bb4bedf3 ("ipv4: Cache output routes in fib_info nexthops.")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new skbuff flag to allow protocols to confirm neighbour.
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
Add sock_confirm_neigh() helper to confirm the neighbour and
use it for IPv4, IPv6 and VRF before dst_neigh_output.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new sock flag to allow sockets to confirm neighbour.
When same struct dst_entry can be used for many different
neighbours we can not use it for pending confirmations.
As not all call paths lock the socket use full word for
the flag.
Add sk_dst_confirm as replacement for dst_confirm when
called for received packets.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the BCM74371 PHY ID to the list of supported chips. This is a 28nm
technology Gigabit PHY SoC.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When PSAMPLE is a loadable module, spectrum must not be built-in:
drivers/net/built-in.o: In function `mlxsw_sp_rx_listener_sample_func':
spectrum.c:(.text+0xe357e): undefined reference to `psample_sample_packet'
This adds a Kconfig dependency to enforce usable configurations.
Fixes: 98d0f7b9ac ("mlxsw: spectrum: Add packet sample offloading support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Yotam Gigi <yotamg@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes the following sparse warnings:
net/ipv6/seg6_iptunnel.c:58:5: warning:
symbol 'nla_put_srh' was not declared. Should it be static?
net/ipv6/seg6_iptunnel.c:238:5: warning:
symbol 'seg6_input' was not declared. Should it be static?
net/ipv6/seg6_iptunnel.c:254:5: warning:
symbol 'seg6_output' was not declared. Should it be static?
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove .owner field if calls are used which set it automatically.
Generated by: scripts/coccinelle/api/platform_no_drv_owner.cocci
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
module_spi_driver() makes the code simpler by eliminating
boilerplate code.
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli says:
====================
net: dsa: Support for pdata in dsa2
This is not exactly new, and was sent before, although back then, I did not
have an user of the pre-declared MDIO board information, but now we do. Note
that I have additional changes queued up to have b53 register platform data for
MIPS bcm47xx and bcm63xx.
Yes I know that we should have the Orion platforms eventually be converted to
Device Tree, but until that happens, I don't want any remaining users of the
old "dsa" platform device (hence the previous DTS submissions for ARM/mvebu)
and, there will be platforms out there that most likely won't never see DT
coming their way (BCM47xx is almost 100% sure, BCM63xx maybe not in a distant
future).
We would probably want the whole series to be merged via David Miller's tree
to simplify things.
Thanks!
Changes in v5:
- dropped changes to drivers/base/ because after more than a month, we cannot
get any answer from Greg KH
Changes in v4:
- Changed device_find_class() to device_find_in_class_name()
- Added kerneldoc above device_find_in_class_name() to explain what it does
and the calling convention regarding device reference counts
- Changed dev_to_net_device to device_to_net_device() added comments
about what it does and the caller conventions regarding reference counts
Changes in v3:
- Tested EPROBE_DEFER from a mockup MDIO/DSA switch driver and everything
is fine, once the driver finally probes we have access to platform data
as expected
- added comment above dsa_port_is_valid() that port->name is mandatory
for platform data cases
- added an extra check in dsa_parse_member() for a NULL pdata pointer
- fixed a bunch of checkpatch errors and warnings
Changes in v2:
- Rebased against latest net-next/master
- Moved dev_find_class() to device_find_class() into drivers/base/core.c
- Moved dev_to_net_device into net/core/dev.c
- Utilize dsa_chip_data directly instead of dsa_platform_data
- Augmented dsa_chip_data to be multi-CPU port ready
Changes from last submission (few months back):
- rebased against latest net-next
- do not introduce dsa2_platform_data which was overkill and was meant to
allow us to do exaclty the same things with platform data and Device Tree
we use the existing dsa_platform_data instead
- properly register MDIO devices when the MDIO bus is registered and associate
platform_data with them
- add a change to the Orion platform code to demonstrate how this can be used
Thank you
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Utilize the ability to pass board specific MDIO bus information towards a
particular MDIO device thus allowing us to provide the per-port switch layout
to the Marvell 88E6XXX switch driver.
Since we would end-up with conflicting registration paths, do not register the
"dsa" platform device anymore.
Note that the MDIO devices registered by code in net/dsa/dsa2.c does not
parse a dsa_platform_data, but directly take a dsa_chip_data (specific
to a single switch chip), so we update the different call sites to pass
this structure down to orion_ge00_switch_init().
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow board support code to collect pre-declarations for MDIO devices by
registering them with mdiobus_register_board_info(). SPI and I2C buses
have a similar feature, we were missing this for MDIO devices, but this
is particularly useful for e.g: MDIO-connected switches which need to
provide their port layout (often board-specific) to a MDIO Ethernet
switch driver.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow drivers to use the new DSA API with platform data. Most of the
code in net/dsa/dsa2.c does not rely so much on device_nodes and can get
the same information from platform_data instead.
We purposely do not support distributed configurations with platform
data, so drivers should be providing a pointer to a 'struct
dsa_chip_data' structure if they wish to communicate per-port layout.
Multiple CPUs port could potentially be supported and dsa_chip_data is
extended to receive up to one reference to an upstream network device
per port described by a dsa_chip_data structure.
dsa_dev_to_net_device() increments the network device's reference count,
so we intentionally call dev_put() to be consistent with the DT-enabled
path, until we have a generic notifier based solution.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation for using this function in net/dsa/dsa2.c, rename the function
to make its scope DSA specific, and export it.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the remaining port configuration code which varies per device
into port.c, using ops were necessary. This makes
mv88e6xxx_6185_family() and mv88e6xxx_6095_family() unused, so remove
them.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mv88e6390 MDIO bus controllers can support for clause 45 accesses.
The internal SERDES interfaces need this, and it is likely external
10GHz PHYs will be clause 45.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Andrew Lunn says:
====================
Set the CMODE for mv88e6390 ports
The mv88e6390 ports 9 & 10 allow there CMODE to be set. CMODE is part
of what linux defines as phy-mode. Add the needed phy-modes to linux,
and add code which will act upon the phy-mode property to configure
the switch port.
These patches have been posted before as part of a bigger patchset
which has now been broken up. I've added the received reviewed by
tags, and added device tree documentation.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Unlike most ports, ports 9 and 10 of the 6390X family have configurable
PHY modes. Set the mode as part of adjust_link().
Ordering is important, because the SERDES interfaces connected to
ports 9 and 10 can be split and assigned to other ports. The CMODE has
to be correctly set before the SERDES interface on another port can be
configured. Such configuration is likely to be performed in
port_enable() and port_disabled(), called on slave_open() and
slave_close().
The simple case is port 9 and 10 are used for 'CPU' or 'DSA'. In this
case, the CMODE is set via a phy-mode in dsa_cpu_dsa_setup(), which is
called early in the switch setup.
When ports 9 or 10 are used as user ports, and have a fixed-phy, when
the fixed fixed-phy is attached, dsa_slave_adjust_link() is called,
which results in the adjust_link function being called, setting the
cmode. The port_enable() will for other ports will be called much
later.
When ports 9 or 10 are used as user ports and have a real phy attached
which does not use all the available SERDES interface, e.g. a 1Gbps
SGMII, there is currently no mechanism in place to set the CMODE of
the port from software. It must be hoped the stripping resistors are
correct.
At the same time, add a function to get the cmode. This will be needed
when configuring the SERDES interfaces.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mv88e6390 ports 9 and 10 supports some additional PHY modes. Add
these modes to the PHY core so they can be used in the binding.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
John Fastabend says:
====================
XDP adjust head support for virtio
This series adds adjust head support for virtio. The following is my
test setup. I use qemu + virtio as follows,
./x86_64-softmmu/qemu-system-x86_64 \
-hda /var/lib/libvirt/images/Fedora-test0.img \
-m 4096 -enable-kvm -smp 2 -netdev tap,id=hn0,queues=4,vhost=on \
-device virtio-net-pci,netdev=hn0,mq=on,guest_tso4=off,guest_tso6=off,guest_ecn=off,guest_ufo=off,vectors=9
In order to use XDP with virtio until LRO is supported TSO must be
turned off in the host. The important fields in the above command line
are the following,
guest_tso4=off,guest_tso6=off,guest_ecn=off,guest_ufo=off
Also note it is possible to conusme more queues than can be supported
because when XDP is enabled for retransmit XDP attempts to use a queue
per cpu. My standard queue count is 'queues=4'.
After loading the VM I run the relevant XDP test programs in,
./sammples/bpf
For this series I tested xdp1, xdp2, and xdp_tx_iptunnel. I usually test
with iperf (-d option to get bidirectional traffic), ping, and pktgen.
I also have a modified xdp1 that returns XDP_PASS on any packet to ensure
the normal traffic path to the stack continues to work with XDP loaded.
It would be great to automate this soon. At the moment I do it by hand
which is starting to get tedious.
v2: original series dropped trace points after merge.
====================
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for XDP adjust head by allocating a 256B header region
that XDP programs can grow into. This is only enabled when a XDP
program is loaded.
In order to ensure that we do not have to unwind queue headroom push
queue setup below bpf_prog_add. It reads better to do a prog ref
unwind vs another queue setup call.
At the moment this code must do a full reset to ensure old buffers
without headroom on program add or with headroom on program removal
are not used incorrectly in the datapath. Ideally we would only
have to disable/enable the RX queues being updated but there is no
API to do this at the moment in virtio so use the big hammer. In
practice it is likely not that big of a problem as this will only
happen when XDP is enabled/disabled changing programs does not
require the reset. There is some risk that the driver may either
have an allocation failure or for some reason fail to correctly
negotiate with the underlying backend in this case the driver will
be left uninitialized. I have not seen this ever happen on my test
systems and for what its worth this same failure case can occur
from probe and other contexts in virtio framework.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For XDP we will need to reset the queues to allow for buffer headroom
to be configured. In order to do this we need to essentially run the
freeze()/restore() code path. Unfortunately the locking requirements
between the freeze/restore and reset paths are different however so
we can not simply reuse the code.
This patch refactors the code path and adds a reset helper routine.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Factor out qp assignment.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
At this point the do_xdp_prog is mostly if/else branches handling
the different modes of virtio_net. So remove it and handle running
the program in the per mode handlers.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For XDP use case and to allow ethtool reset tests it is useful to be
able to use reset paths from contexts where rtnl lock is already
held.
This requries updating virtnet_set_queues and free_receive_bufs the
two places where rtnl_lock is taken in virtio_net. To do this we
use the following pattern,
_foo(...) { do stuff }
foo(...) { rtnl_lock(); _foo(...); rtnl_unlock()};
this allows us to use freeze()/restore() flow from both contexts.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Nikolay Aleksandrov says:
====================
bridge: improve cache utilization
This is the first set which begins to deal with the bad bridge cache
access patterns. The first patch rearranges the bridge and port structs
a little so the frequently (and closely) accessed members are in the same
cache line. The second patch then moves the garbage collection to a
workqueue trying to improve system responsiveness under load (many fdbs)
and more importantly removes the need to check if the matched entry is
expired in __br_fdb_get which was a major source of false-sharing.
The third patch is a preparation for the final one which
If properly configured, i.e. ports bound to CPUs (thus updating "updated"
locally) then the bridge's HitM goes from 100% to 0%, but even without
binding we get a win because previously every lookup that iterated over
the hash chain caused false-sharing due to the first cache line being
used for both mac/vid and used/updated fields.
Some results from tests I've run:
(note that these were run in good conditions for the baseline, everything
ran on a single NUMA node and there were only 3 fdbs)
1. baseline
100% Load HitM on the fdbs (between everyone who has done lookups and hit
one of the 3 hash chains of the communicating
src/dst fdbs)
Overall 5.06% Load HitM for the bridge, first place in the list
2. patched & ports bound to CPUs
0% Local load HitM, bridge is not even in the c2c report list
Also there's 3% consistent improvement in netperf tests.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Writing once per jiffy is enough to limit the bridge's false sharing.
After this change the bridge doesn't show up in the local load HitM stats.
Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fdb's used and updated fields are written to on every packet forward and
packet receive respectively. Thus if we are receiving packets from a
particular fdb, they'll cause false-sharing with everyone who has looked
it up (even if it didn't match, since mac/vid share cache line!). The
"used" field is even worse since it is updated on every packet forward
to that fdb, thus the standard config where X ports use a single gateway
results in 100% fdb false-sharing. Note that this patch does not prevent
the last scenario, but it makes it better for other bridge participants
which are not using that fdb (and are only doing lookups over it).
The point is with this move we make sure that only communicating parties
get the false-sharing, in a later patch we'll show how to avoid that too.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the fdb garbage collector to a workqueue which fires at least 10
milliseconds apart and cleans chain by chain allowing for other tasks
to run in the meantime. When having thousands of fdbs the system is much
more responsive. Most importantly remove the need to check if the
matched entry has expired in __br_fdb_get that causes false-sharing and
is completely unnecessary if we cleanup entries, at worst we'll get 10ms
of traffic for that entry before it gets deleted.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move around net_bridge so the vlan fields are in the beginning since
they're checked on every packet even if vlan filtering is disabled.
For the port move flags & vlan group to the beginning, so they're in the
same cache line with the port's state (both flags and state are checked
on each packet).
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The patch fixes the case when adding a zero value to the packet
pointer. The zero value could come from src_reg equals type
BPF_K or CONST_IMM. The patch fixes both, otherwise the verifer
reports the following error:
[...]
R0=imm0,min_value=0,max_value=0
R1=pkt(id=0,off=0,r=4)
R2=pkt_end R3=fp-12
R4=imm4,min_value=4,max_value=4
R5=pkt(id=0,off=4,r=4)
269: (bf) r2 = r0 // r2 becomes imm0
270: (77) r2 >>= 3
271: (bf) r4 = r1 // r4 becomes pkt ptr
272: (0f) r4 += r2 // r4 += 0
addition of negative constant to packet pointer is not allowed
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
These two tests are based on the work done for f23cc643f9. The first test is
just a basic one to make sure we don't allow AND'ing negative values, even if it
would result in a valid index for the array. The second is a cleaned up version
of the original testcase provided by Jann Horn that resulted in the commit.
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot says:
====================
net: dsa: add fabric notifier
When a switch fabric is composed of multiple switch chips, these chips
must be programmed accordingly when an event occurred on one of them.
Examples of such event include hardware bridging: when a Linux bridge
spans interconnected chips, they must be programmed to allow external
ports to ingress frames on their internal ports.
Another example is cross-chip hardware VLANs. Switch chips in-between
interconnected bridge ports must also configure a given VLAN to allow
packets to pass through them.
In order to support that, this patchset introduces a non-intrusive
notifier mechanism. It adds a notifier head in every DSA switch tree
(the said fabric), and a notifier block in every DSA switch chip.
When an even occurs, it is chained to all notifiers of the fabric.
Switch chips can react accordingly if they are cross-chip capable.
On a dynamic debug enabled system, bridging a port in a multi-chip
fabric will print something like this (ZII Rev B board):
# brctl addif br0 lan3
mv88e6085 0.1:00: crosschip DSA port 1.0 bridged to br0
mv88e6085 0.4:00: crosschip DSA port 1.0 bridged to br0
# brctl delif br0 lan3
mv88e6085 0.1:00: crosschip DSA port 1.0 unbridged from br0
mv88e6085 0.4:00: crosschip DSA port 1.0 unbridged from br0
Currently only bridging events are added. A patchset introducing support
for cross-chip hardware bridging configuration in mv88e6xxx will follow
right after. Then events for switchdev operations are next on the line.
We should note that non-switchdev events do not support rolling-back
switch-wide operations. We'll have to work on closer integration with
switchdev for that, like introducing new attributes or objects, to
benefit from the prepare and commit phases.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
A slave device will now notify the switch fabric once its port is
bridged or unbridged, instead of calling directly its switch operations.
This code allows propagating cross-chip bridging events in the fabric.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
