Eric Dumazet [Tue, 15 Nov 2016 18:15:11 +0000 (10:15 -0800)]
net: busy-poll: allow preemption in sk_busy_loop()
After commit 4cd13c21b207 ("softirq: Let ksoftirqd do its job"),
sk_busy_loop() needs a bit of care :
softirqs might be delayed since we do not allow preemption yet.
This patch adds preemptiom points in sk_busy_loop(),
and makes sure no unnecessary cache line dirtying
or atomic operations are done while looping.
A new flag is added into napi->state : NAPI_STATE_IN_BUSY_POLL
This prevents napi_complete_done() from clearing NAPIF_STATE_SCHED,
so that sk_busy_loop() does not have to grab it again.
Similarly, netpoll_poll_lock() is done one time.
This gives about 10 to 20 % improvement in various busy polling
tests, especially when many threads are busy polling in
configurations with large number of NIC queues.
This should allow experimenting with bigger delays without
hurting overall latencies.
Tested:
On a 40Gb mlx4 NIC, 32 RX/TX queues.
echo 70 >/proc/sys/net/core/busy_read
for i in `seq 1 40`; do echo -n $i: ; ./super_netperf $i -H lpaa24 -t UDP_RR -- -N -n; done
Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Willem de Bruijn <willemb@google.com> Cc: Adam Belay <abelay@google.com> Cc: Tariq Toukan <tariqt@mellanox.com> Cc: Yuval Mintz <Yuval.Mintz@cavium.com> Cc: Ariel Elior <ariel.elior@cavium.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Tue, 15 Nov 2016 19:00:04 +0000 (11:00 -0800)]
bpf: Fix compilation warning in __bpf_lru_list_rotate_inactive
gcc-6.2.1 gives the following warning:
kernel/bpf/bpf_lru_list.c: In function ‘__bpf_lru_list_rotate_inactive.isra.3’:
kernel/bpf/bpf_lru_list.c:201:28: warning: ‘next’ may be used uninitialized in this function [-Wmaybe-uninitialized]
The "next" is currently initialized in the while() loop which must have >=1
iterations.
This patch initializes next to get rid of the compiler warning.
Fixes: 3a08c2fd7634 ("bpf: LRU List") Reported-by: David Miller <davem@davemloft.net> Signed-off-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
David Lebrun [Tue, 15 Nov 2016 15:14:04 +0000 (16:14 +0100)]
ipv6: sr: add option to control lwtunnel support
This patch adds a new option CONFIG_IPV6_SEG6_LWTUNNEL to enable/disable
support of encapsulation with the lightweight tunnels. When this option
is enabled, CONFIG_LWTUNNEL is automatically selected.
Fix commit 6c8702c60b88 ("ipv6: sr: add support for SRH encapsulation and injection with lwtunnels")
Without a proper option to control lwtunnel support for SR-IPv6, if
CONFIG_LWTUNNEL=n then the IPv6 initialization fails as a consequence
of seg6_iptunnel_init() failure with EOPNOTSUPP:
NET: Registered protocol family 10
IPv6: Attempt to unregister permanent protocol 6
IPv6: Attempt to unregister permanent protocol 136
IPv6: Attempt to unregister permanent protocol 17
NET: Unregistered protocol family 10
Tested (compiling, booting, and loading ipv6 module when relevant)
with possible combinations of CONFIG_IPV6={y,m,n},
CONFIG_IPV6_SEG6_LWTUNNEL={y,n} and CONFIG_LWTUNNEL={y,n}.
Reported-by: Lorenzo Colitti <lorenzo@google.com> Suggested-by: Roopa Prabhu <roopa@cumulusnetworks.com> Signed-off-by: David Lebrun <david.lebrun@uclouvain.be> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 16 Nov 2016 03:46:31 +0000 (22:46 -0500)]
Merge branch 'alx-multiqueue-support'
Tobias Regnery says:
====================
alx: add multi queue support
This patchset lays the groundwork for multi queue support in the alx driver
and enables multi queue support for the tx path by default. The hardware
supports up to 4 tx queues.
Benefits are better utilization of multi core cpus and the usage of the
msi-x support by default which splits the handling of rx / tx and misc
other interrupts.
The rx path is a little bit harder because apparently (based on the limited
information from the downstream driver) the hardware supports up to 8 rss
queues but only has one hardware descriptor ring on the rx side. So the rx
path will be part of another patchset.
Tested on my AR8161 ethernet adapter with different tests:
- there are no regressions observed during my daily usage
- iperf tcp and udp tests shows no performance regressions
- netperf TCP_RR and UDP_RR shows a slight performance increase of about
1-2% with this patchset applied
This work is based on the downstream driver at github.com/qca/alx
Changes in V2:
- drop unneeded casts in alx_alloc_rx_ring (Patch 1)
- add additional information about testing and benefit to the
changelog
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Regnery [Tue, 15 Nov 2016 11:43:15 +0000 (12:43 +0100)]
alx: enable msi-x interrupts by default
Remove the module parameter to enable msi-x support and enable msi-x
interrupts unconditionally by default. This is a preparatory step to enable
multi queue support by default, because this is only working with msi-x
interrupts.
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Regnery [Tue, 15 Nov 2016 11:43:14 +0000 (12:43 +0100)]
alx: prepare tx path for multi queue support
This patch prepares the tx path to send data on multiple tx queues. It
introduces per queue register adresses and uses them in the alx_tx_queue
structs.
There are new helper functions for the queue mapping in the tx path.
Based on the downstream driver at github.com/qca/alx
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Regnery [Tue, 15 Nov 2016 11:43:13 +0000 (12:43 +0100)]
alx: prepare resource allocation for multi queue support
Allocate, initialise and free alx_tx_queue structs based on the number of
alx_napi structures. Also increase the size of the descriptor memory based
on the number of tx queues in use.
Based on the downstream driver at github.com/qca/alx
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Regnery [Tue, 15 Nov 2016 11:43:10 +0000 (12:43 +0100)]
alx: add ability to allocate and free alx_napi structures
Add new functions to allocate and free the alx_napi structures and use them
in __alx_open and __alx_stop. We only allocate one of these structures for
now, as the rest of the driver is not yet ready for multiple queues.
We switch over the setup of the interrupt mask and the call to netif_napi_add
to the new function because we must adjust these later on a per queue basis.
Based on the downstream driver at github.com/qca/alx
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Regnery [Tue, 15 Nov 2016 11:43:08 +0000 (12:43 +0100)]
alx: refactor descriptor allocation
Split the allocation of descriptor memory and the buffer allocation into a
tx and rx function. This is in preparation for multiple queues where we
need to iterate over the new functions.
While at it drop the unneeded casting on the rx side.
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 16 Nov 2016 03:34:27 +0000 (22:34 -0500)]
Merge branch 'dpaa_eth-next'
Madalin Bucur says:
====================
dpaa_eth: Add the QorIQ DPAA Ethernet driver
This patch series adds the Ethernet driver for the Freescale
QorIQ Data Path Acceleration Architecture (DPAA).
This version includes changes following the feedback received
on previous versions from Eric Dumazet, Bob Cochran, Joe Perches,
Paul Bolle, Joakim Tjernlund, Scott Wood, David Miller - thank you.
Together with the driver a managed version of alloc_percpu
is provided that simplifies the release of per-CPU memory.
The Freescale DPAA architecture consists in a series of hardware
blocks that support the Ethernet connectivity. The Ethernet driver
depends upon the following drivers that are currently in the Linux
kernel:
- Peripheral Access Memory Unit (PAMU)
drivers/iommu/fsl_*
- Frame Manager (FMan) added in v4.4
drivers/net/ethernet/freescale/fman
- Queue Manager (QMan), Buffer Manager (BMan) added in v4.9-rc1
drivers/soc/fsl/qbman
where the acronyms used above (and in the code) are:
DPAA = Data Path Acceleration Architecture
FMan = DPAA Frame Manager
QMan = DPAA Queue Manager
BMan = DPAA Buffers Manager
QMI = QMan interface in FMan
BMI = BMan interface in FMan
FMan SP = FMan Storage Profiles
MURAM = Multi-user RAM in FMan
FQ = QMan Frame Queue
Rx Dfl FQ = default reception FQ
Rx Err FQ = Rx error frames FQ
Tx Cnf FQ = Tx confirmation FQ
Tx FQs = transmission frame queues
dtsec = datapath three speed Ethernet controller (10/100/1000 Mbps)
tgec = ten gigabit Ethernet controller (10 Gbps)
memac = multirate Ethernet MAC (10/100/1000/10000)
Changes from v7:
- remove the debug option to use a common buffer pool for all the
interfaces
Changed from v6:
- fixed an issue on an error path in dpaa_set_mac_address()
- removed NDO operation definitions that were not needed
- sorted the local variable declarations
- cleaned up a few checkpatch checks
- removed friendly network interface naming code
Changes from v5:
- adapt to the latest Q/BMan drivers API
- use build_skb() on Rx path instead of buffer pool refill path
- proper support for multiple buffer pools
- align function, variable names, code cleanup
- driver file structure cleanup
Changes from v4:
- addressed feedback from Scott Wood and Joe Perches
- fixed spelling
- fixed leak of uninitialized stack to userspace
- fix prints
- replace raw_cpu_ptr() with this_cpu_ptr()
- remove _s from the end of structure names
- remove underscores at start of functions, goto labels
- remove likely in error paths
- use container_of() instead of open casts
- remove priv from the driver name
- move return type on same line with function name
- drop DPA_READ_SKB_PTR/DPA_WRITE_SKB_PTR
Changes from v3:
- removed bogus delay and comment in .ndo_stop implementation
- addressed minor issues reported by David Miller
Changes from v2:
- removed debugfs, moved exports to ethtool statistics
- removed congestion groups Kconfig params
Changes from v1:
- bpool level Kconfig options removed
- print format using pr_fmt, cleaned up prints
- __hot/__cold removed
- gratuitous unlikely() removed
- code style aligned, consistent spacing for declarations
- comment formatting
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Madalin Bucur [Tue, 15 Nov 2016 08:41:04 +0000 (10:41 +0200)]
dpaa_eth: add ethtool statistics
Add a series of counters to be exported through ethtool:
- add detailed counters for reception errors;
- add detailed counters for QMan enqueue reject events;
- count the number of fragmented skbs received from the stack;
- count all frames received on the Tx confirmation path;
- add congestion group statistics;
- count the number of interrupts for each CPU.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Madalin Bucur [Tue, 15 Nov 2016 08:41:02 +0000 (10:41 +0200)]
dpaa_eth: add support for DPAA Ethernet
This introduces the Freescale Data Path Acceleration Architecture
(DPAA) Ethernet driver (dpaa_eth) that builds upon the DPAA QMan,
BMan, PAMU and FMan drivers to deliver Ethernet connectivity on
the Freescale DPAA QorIQ platforms.
Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Andrey Vagin [Tue, 15 Nov 2016 02:15:14 +0000 (18:15 -0800)]
tcp: allow to enable the repair mode for non-listening sockets
The repair mode is used to get and restore sequence numbers and
data from queues. It used to checkpoint/restore connections.
Currently the repair mode can be enabled for sockets in the established
and closed states, but for other states we have to dump the same socket
properties, so lets allow to enable repair mode for these sockets.
The repair mode reveals nothing more for sockets in other states.
Signed-off-by: Andrei Vagin <avagin@openvz.org> Acked-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 16 Nov 2016 03:24:42 +0000 (22:24 -0500)]
Merge branch 'liquidio-CN23XX-VF-support'
Raghu Vatsavayi says:
====================
liquidio CN23XX VF support
Following is the V6 patch series for adding VF support on
CN23XX devices. This version addressed:
1) Your concern for ordering of local variable declarations
from longest to shortest line.
2) Removed module parameters max_vfs, num_queues_per_{p,v}f.
3) Minor changes for fixing new checkpatch script related
errors on pre-existing driver.
4) Fixed compilation issues when CONFIG_PCI_IOV/CONFIG_PCI_ATS
options are disabled.
5) Modified qualifiers for printing mac addresses with pM format.
I will post remaining VF patches soon after this patchseries is
applied. Please apply patches in the following order as some of
the patches depend on earlier patches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Lendacky, Thomas [Mon, 14 Nov 2016 22:39:16 +0000 (16:39 -0600)]
amd-xgbe: Fix up some coccinelle identified warnings
Fix up some warnings that were identified by coccinelle:
Clean up an if/else block that can look confusing since the same statement
is executed in an "else if" check and the final "else" statement.
Change a variable from unsigned int to int since it is used in an if
statement checking the value to be less than 0.
Reported-by: Julia Lawall <julia.lawall@lip6.fr> Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Lendacky, Thomas [Mon, 14 Nov 2016 22:39:07 +0000 (16:39 -0600)]
amd-xgbe: Fix mask appliciation for Clause 37 register
The application of a mask to clear an area of a clause 37 register value
was not properly applied. Update the code to do the proper application
of the mask.
Reported-by: Marion & Christophe JAILLET <christophe.jaillet@wanadoo.fr> Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 16 Nov 2016 03:05:55 +0000 (22:05 -0500)]
Merge branch 'sun4i-emac-big-endian'
Michael Weiser says:
====================
sun4i-emac: Fixes for running a big-endian kernel on Cubieboard2
the following patches are what remains to be fixed in order to allow running a
big-endian kernel on the Cubieboard2.
The first patch fixes up endianness problems with DMA descriptors in
the stmmac driver preventing it from working correctly when runnning a
big-endian kernel.
The second patch adds the ability to enable diagnostic messages in the
sun4i-emac driver which were instrumental in finding the problem fixed
by patch number three: Endianness confusion caused by dual-purpose I/O
register usage in sun4i-emac.
All of these have been tested successfully on a Cubieboard2 DualCard.
Changes since v4:
- Rebased to current master
- Removed already applied patches to sunxi-mmc and sunxi-Kconfig
Changes since v3:
- Rebased sunxi-mmc patch against Ulf's mmc.git/next
- Changed Kconfig change to enable big-endian support only for sun7i
devices
Changes since v2:
- Fixed typo in stmmac patch causing a build failure
- Added sun4i-emac patches
Michael Weiser [Mon, 14 Nov 2016 17:58:07 +0000 (18:58 +0100)]
net: ethernet: sun4i-emac: Read rxhdr in CPU byte-order
The EMAC EMAC_RX_IO_DATA_REG data register is dual-purpose: On one hand
it is used to move actual packet data off the wire. This will be in
wire-format and accepted as such by higher layers such as IP. Therefore
it is correctly read as-is (i.e. raw) using readsl.
On the other hand it provides metadata about incoming transfers to the
driver such as length and checksum validation status. This data is
little-endian, always and it is interpreted by the driver. Therefore it
needs to be swapped to CPU endianness to make sense to the driver. This
is already done for the "receive header" but not rxhdr.
Read rxhdr using readl in order for sun4i-emac to work correctly when
running a big-endian kernel.
Signed-off-by: Michael Weiser <michael.weiser@gmx.de> Cc: Maxime Ripard <maxime.ripard@free-electrons.com> Cc: netdev@vger.kernel.org Signed-off-by: David S. Miller <davem@davemloft.net>
Michael Weiser [Mon, 14 Nov 2016 17:58:06 +0000 (18:58 +0100)]
net: ethernet: sun4i-emac: Allow to enable netif messages
sun4i-emac has the ability to print a number of diagnostic messages using
dev_dbg depending on message level settings implemented using netif_msg_*
macros. But there's no way to actually enable them.
Add the ability to switch diagnostic messages on using either a module
parameter debug or ethtool -s <netif> msglvl <flags>.
Signed-off-by: Michael Weiser <michael.weiser@gmx.de> Cc: Maxime Ripard <maxime.ripard@free-electrons.com> Cc: netdev@vger.kernel.org Signed-off-by: David S. Miller <davem@davemloft.net>
Michael Weiser [Mon, 14 Nov 2016 17:58:05 +0000 (18:58 +0100)]
net: ethernet: stmmac: change dma descriptors to __le32
The stmmac driver does not take into account the processor may be big
endian when writing the DMA descriptors. This causes the ethernet
interface not to be initialised correctly when running a big-endian
kernel. Change the descriptors for DMA to use __le32 and ensure they are
suitably swapped before writing. Tested successfully on the
Cubieboard2.
Signed-off-by: Michael Weiser <michael.weiser@gmx.de> Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com> Cc: Alexandre Torgue <alexandre.torgue@st.com> Cc: netdev@vger.kernel.org Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Westphal [Mon, 14 Nov 2016 15:42:01 +0000 (16:42 +0100)]
dctcp: update cwnd on congestion event
draft-ietf-tcpm-dctcp-02 says:
... when the sender receives an indication of congestion
(ECE), the sender SHOULD update cwnd as follows:
cwnd = cwnd * (1 - DCTCP.Alpha / 2)
So, lets do this and reduce cwnd more smoothly (and faster), as per
current congestion estimate.
Cc: Lawrence Brakmo <brakmo@fb.com> Cc: Andrew Shewmaker <agshew@gmail.com> Cc: Glenn Judd <glenn.judd@morganstanley.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Wed, 16 Nov 2016 02:21:09 +0000 (18:21 -0800)]
net: bcm63xx_enet: Fix build failure with phy_ethtool_nway_reset
Introduced a typo making the driver no longer build, *sigh*.
Fixes: 42469bf5d9bb ("net: bcm63xx_enet: Utilize phy_ethtool_nway_reset") Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
This patch series centralizes how ethtool::nway_reset is implemented
by providing a PHYLIB function which calls into genphy_restart_aneg().
All drivers below are converted to use this new helper function. Some
other have specific requirements that make them not quite suitable for
a straight forward conversion.
There is another patch series which implements ethtool::nway_reset
using the helper function introduced that depends on this patch
series.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Tue, 15 Nov 2016 18:06:44 +0000 (10:06 -0800)]
net: usb: lan78xx: Utilize phy_ethtool_nway_reset
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Acked-by: Woojung Huh <woojung.huh@microchip.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Reviewed-by: Thomas Langer <Thomas.langer@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Tue, 15 Nov 2016 21:32:11 +0000 (16:32 -0500)]
vxlan: Fix uninitialized variable warnings.
drivers/net/vxlan.c: In function ‘vxlan_xmit_one’:
drivers/net/vxlan.c:2141:10: warning: ‘err’ may be used uninitialized in this function [-Wmaybe-uninitialized]
Signed-off-by: David S. Miller <davem@davemloft.net>
pravin shelar [Mon, 14 Nov 2016 04:43:57 +0000 (20:43 -0800)]
vxlan: simplify vxlan xmit
Existing vxlan xmit function handles two distinct cases.
1. vxlan net device
2. vxlan lwt device.
By seperating initialization these two cases the egress path
looks better.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org> Acked-by: Jiri Benc <jbenc@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
pravin shelar [Mon, 14 Nov 2016 04:43:55 +0000 (20:43 -0800)]
vxlan: improve vxlan route lookup checks.
Move route sanity check to respective vxlan[4/6]_get_route functions.
This allows us to perform all sanity checks before caching the dst so
that we can avoid these checks on subsequent packets.
This give move accurate metadata information for packet from
fill_metadata_dst().
Signed-off-by: Pravin B Shelar <pshelar@ovn.org> Acked-by: Jiri Benc <jbenc@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
pravin shelar [Mon, 14 Nov 2016 04:43:54 +0000 (20:43 -0800)]
vxlan: simplify exception handling
vxlan egress path error handling has became complicated, it
need to handle IPv4 and IPv6 tunnel cases.
Earlier patch removes vlan handling from vxlan_build_skb(), so
vxlan_build_skb does not need to free skb and we can simplify
the xmit path by having single error handling for both type of
tunnels.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org> Acked-by: Jiri Benc <jbenc@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
pravin shelar [Mon, 14 Nov 2016 04:43:52 +0000 (20:43 -0800)]
vxlan: avoid vlan processing in vxlan device.
VxLan device does not have special handling for vlan taging on egress.
Therefore it does not make sense to expose vlan offloading feature.
This patch does not change vxlan functinality.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org> Acked-by: Jiri Benc <jbenc@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Paolo Abeni [Tue, 15 Nov 2016 15:37:53 +0000 (16:37 +0100)]
udplite: fix NULL pointer dereference
The commit 850cbaddb52d ("udp: use it's own memory accounting schema")
assumes that the socket proto has memory accounting enabled,
but this is not the case for UDPLITE.
Fix it enabling memory accounting for UDPLITE and performing
fwd allocated memory reclaiming on socket shutdown.
UDP and UDPLITE share now the same memory accounting limits.
Also drop the backlog receive operation, since is no more needed.
Fixes: 850cbaddb52d ("udp: use it's own memory accounting schema") Reported-by: Andrei Vagin <avagin@gmail.com> Suggested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Fri, 11 Nov 2016 18:55:11 +0000 (10:55 -0800)]
bpf: Add tests for the LRU bpf_htab
This patch has some unit tests and a test_lru_dist.
The test_lru_dist reads in the numeric keys from a file.
The files used here are generated by a modified fio-genzipf tool
originated from the fio test suit. The sample data file can be
found here: https://github.com/iamkafai/bpf-lru
The zipf.* data files have 100k numeric keys and the key is also
ranged from 1 to 100k.
The test_lru_dist outputs the number of unique keys (nr_unique).
F.e. The following means, 61239 of them is unique out of 100k keys.
nr_misses means it cannot be found in the LRU map, so nr_misses
must be >= nr_unique. test_lru_dist also simulates a perfect LRU
map as a comparison:
Martin KaFai Lau [Fri, 11 Nov 2016 18:55:07 +0000 (10:55 -0800)]
bpf: Add percpu LRU list
Instead of having a common LRU list, this patch allows a
percpu LRU list which can be selected by specifying a map
attribute. The map attribute will be added in the later
patch.
While the common use case for LRU is #reads >> #updates,
percpu LRU list allows bpf prog to absorb unusual #updates
under pathological case (e.g. external traffic facing machine which
could be under attack).
Each percpu LRU is isolated from each other. The LRU nodes (including
free nodes) cannot be moved across different LRU Lists.
Here are the update performance comparison between
common LRU list and percpu LRU list (the test code is
at the last patch):
[root@kerneltest003.31.prn1 ~]# for i in 1 4 8; do echo -n "$i cpus: "; \
./map_perf_test 16 $i | awk '{r += $3}END{print r " updates"}'; done
1 cpus: 2934082 updates
4 cpus: 7391434 updates
8 cpus: 6500576 updates
[root@kerneltest003.31.prn1 ~]# for i in 1 4 8; do echo -n "$i cpus: "; \
./map_perf_test 32 $i | awk '{r += $3}END{printr " updates"}'; done
1 cpus: 2896553 updates
4 cpus: 9766395 updates
8 cpus: 17460553 updates
Signed-off-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Fri, 11 Nov 2016 18:55:06 +0000 (10:55 -0800)]
bpf: LRU List
Introduce bpf_lru_list which will provide LRU capability to
the bpf_htab in the later patch.
* General Thoughts:
1. Target use case. Read is more often than update.
(i.e. bpf_lookup_elem() is more often than bpf_update_elem()).
If bpf_prog does a bpf_lookup_elem() first and then an in-place
update, it still counts as a read operation to the LRU list concern.
2. It may be useful to think of it as a LRU cache
3. Optimize the read case
3.1 No lock in read case
3.2 The LRU maintenance is only done during bpf_update_elem()
4. If there is a percpu LRU list, it will lose the system-wise LRU
property. A completely isolated percpu LRU list has the best
performance but the memory utilization is not ideal considering
the work load may be imbalance.
5. Hence, this patch starts the LRU implementation with a global LRU
list with batched operations before accessing the global LRU list.
As a LRU cache, #read >> #update/#insert operations, it will work well.
6. There is a local list (for each cpu) which is named
'struct bpf_lru_locallist'. This local list is not used to sort
the LRU property. Instead, the local list is to batch enough
operations before acquiring the lock of the global LRU list. More
details on this later.
7. In the later patch, it allows a percpu LRU list by specifying a
map-attribute for scalability reason and for use cases that need to
prepare for the worst (and pathological) case like DoS attack.
The percpu LRU list is completely isolated from each other and the
LRU nodes (including free nodes) cannot be moved across the list. The
following description is for the global LRU list but mostly applicable
to the percpu LRU list also.
* Global LRU List:
1. It has three sub-lists: active-list, inactive-list and free-list.
2. The two list idea, active and inactive, is borrowed from the
page cache.
3. All nodes are pre-allocated and all sit at the free-list (of the
global LRU list) at the beginning. The pre-allocation reasoning
is similar to the existing BPF_MAP_TYPE_HASH. However,
opting-out prealloc (BPF_F_NO_PREALLOC) is not supported in
the LRU map.
* Active/Inactive List (of the global LRU list):
1. The active list, as its name says it, maintains the active set of
the nodes. We can think of it as the working set or more frequently
accessed nodes. The access frequency is approximated by a ref-bit.
The ref-bit is set during the bpf_lookup_elem().
2. The inactive list, as its name also says it, maintains a less
active set of nodes. They are the candidates to be removed
from the bpf_htab when we are running out of free nodes.
3. The ordering of these two lists is acting as a rough clock.
The tail of the inactive list is the older nodes and
should be released first if the bpf_htab needs free element.
* Rotating the Active/Inactive List (of the global LRU list):
1. It is the basic operation to maintain the LRU property of
the global list.
2. The active list is only rotated when the inactive list is running
low. This idea is similar to the current page cache.
Inactive running low is currently defined as
"# of inactive < # of active".
3. The active list rotation always starts from the tail. It moves
node without ref-bit set to the head of the inactive list.
It moves node with ref-bit set back to the head of the active
list and then clears its ref-bit.
4. The inactive rotation is pretty simply.
It walks the inactive list and moves the nodes back to the head of
active list if its ref-bit is set. The ref-bit is cleared after moving
to the active list.
If the node does not have ref-bit set, it just leave it as it is
because it is already in the inactive list.
* Shrinking the Inactive List (of the global LRU list):
1. Shrinking is the operation to get free nodes when the bpf_htab is
full.
2. It usually only shrinks the inactive list to get free nodes.
3. During shrinking, it will walk the inactive list from the tail,
delete the nodes without ref-bit set from bpf_htab.
4. If no free node found after step (3), it will forcefully get
one node from the tail of inactive or active list. Forcefully is
in the sense that it ignores the ref-bit.
* Local List:
1. Each CPU has a 'struct bpf_lru_locallist'. The purpose is to
batch enough operations before acquiring the lock of the
global LRU.
2. A local list has two sub-lists, free-list and pending-list.
3. During bpf_update_elem(), it will try to get from the free-list
of (the current CPU local list).
4. If the local free-list is empty, it will acquire from the
global LRU list. The global LRU list can either satisfy it
by its global free-list or by shrinking the global inactive
list. Since we have acquired the global LRU list lock,
it will try to get at most LOCAL_FREE_TARGET elements
to the local free list.
5. When a new element is added to the bpf_htab, it will
first sit at the pending-list (of the local list) first.
The pending-list will be flushed to the global LRU list
when it needs to acquire free nodes from the global list
next time.
* Lock Consideration:
The LRU list has a lock (lru_lock). Each bucket of htab has a
lock (buck_lock). If both locks need to be acquired together,
the lock order is always lru_lock -> buck_lock and this only
happens in the bpf_lru_list.c logic.
In hashtab.c, both locks are not acquired together (i.e. one
lock is always released first before acquiring another lock).
Signed-off-by: Martin KaFai Lau <kafai@fb.com> Acked-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net>
1) Fix off by one wrt. indexing when dumping /proc/net/route entries,
from Alexander Duyck.
2) Fix lockdep splats in iwlwifi, from Johannes Berg.
3) Cure panic when inserting certain netfilter rules when NFT_SET_HASH
is disabled, from Liping Zhang.
4) Memory leak when nft_expr_clone() fails, also from Liping Zhang.
5) Disable UFO when path will apply IPSEC tranformations, from Jakub
Sitnicki.
6) Don't bogusly double cwnd in dctcp module, from Florian Westphal.
7) skb_checksum_help() should never actually use the value "0" for the
resulting checksum, that has a special meaning, use CSUM_MANGLED_0
instead. From Eric Dumazet.
8) Per-tx/rx queue statistic strings are wrong in qed driver, fix from
Yuval MIntz.
9) Fix SCTP reference counting of associations and transports in
sctp_diag. From Xin Long.
10) When we hit ip6tunnel_xmit() we could have come from an ipv4 path in
a previous layer or similar, so explicitly clear the ipv6 control
block in the skb. From Eli Cooper.
11) Fix bogus sleeping inside of inet_wait_for_connect(), from WANG
Cong.
12) Correct deivce ID of T6 adapter in cxgb4 driver, from Hariprasad
Shenai.
13) Fix potential access past the end of the skb page frag array in
tcp_sendmsg(). From Eric Dumazet.
14) 'skb' can legitimately be NULL in inet{,6}_exact_dif_match(). Fix
from David Ahern.
15) Don't return an error in tcp_sendmsg() if we wronte any bytes
successfully, from Eric Dumazet.
16) Extraneous unlocks in netlink_diag_dump(), we removed the locking
but forgot to purge these unlock calls. From Eric Dumazet.
17) Fix memory leak in error path of __genl_register_family(). We leak
the attrbuf, from WANG Cong.
18) cgroupstats netlink policy table is mis-sized, from WANG Cong.
19) Several XDP bug fixes in mlx5, from Saeed Mahameed.
20) Fix several device refcount leaks in network drivers, from Johan
Hovold.
21) icmp6_send() should use skb dst device not skb->dev to determine L3
routing domain. From David Ahern.
22) ip_vs_genl_family sets maxattr incorrectly, from WANG Cong.
23) We leak new macvlan port in some cases of maclan_common_netlink()
errors. Fix from Gao Feng.
24) Similar to the icmp6_send() fix, icmp_route_lookup() should
determine L3 routing domain using skb_dst(skb)->dev not skb->dev.
Also from David Ahern.
25) Several fixes for route offloading and FIB notification handling in
mlxsw driver, from Jiri Pirko.
26) Properly cap __skb_flow_dissect()'s return value, from Eric Dumazet.
27) Fix long standing regression in ipv4 redirect handling, wrt.
validating the new neighbour's reachability. From Stephen Suryaputra
Lin.
28) If sk_filter() trims the packet excessively, handle it reasonably in
tcp input instead of exploding. From Eric Dumazet.
29) Fix handling of napi hash state when copying channels in sfc driver,
from Bert Kenward.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (121 commits)
mlxsw: spectrum_router: Flush FIB tables during fini
net: stmmac: Fix lack of link transition for fixed PHYs
sctp: change sk state only when it has assocs in sctp_shutdown
bnx2: Wait for in-flight DMA to complete at probe stage
Revert "bnx2: Reset device during driver initialization"
ps3_gelic: fix spelling mistake in debug message
net: ethernet: ixp4xx_eth: fix spelling mistake in debug message
ibmvnic: Fix size of debugfs name buffer
ibmvnic: Unmap ibmvnic_statistics structure
sfc: clear napi_hash state when copying channels
mlxsw: spectrum_router: Correctly dump neighbour activity
mlxsw: spectrum: Fix refcount bug on span entries
bnxt_en: Fix VF virtual link state.
bnxt_en: Fix ring arithmetic in bnxt_setup_tc().
Revert "include/uapi/linux/atm_zatm.h: include linux/time.h"
tcp: take care of truncations done by sk_filter()
ipv4: use new_gw for redirect neigh lookup
r8152: Fix error path in open function
net: bpqether.h: remove if_ether.h guard
net: __skb_flow_dissect() must cap its return value
...
* tag 'rtc-4.9-2' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
rtc: omap: prevent disabling of clock/module during suspend
rtc: omap: Fix selecting external osc
rtc: cmos: Don't enable interrupts in the middle of the interrupt handler
rtc: cmos: remove all __exit_p annotations
rtc: asm9260: fix module autoload
Chris Metcalf [Mon, 7 Nov 2016 19:32:02 +0000 (14:32 -0500)]
tile: handle __ro_after_init like parisc does
The tile architecture already marks RO_DATA as read-only in
the kernel, so grouping RO_AFTER_INIT_DATA with RO_DATA, as is
done by default, means the kernel faults in init when it tries
to write to RO_AFTER_INIT_DATA. For now, just arrange that
__ro_after_init is handled like __write_once, i.e. __read_mostly.
Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: Chris Metcalf <cmetcalf@mellanox.com>
Ido Schimmel [Mon, 14 Nov 2016 10:26:32 +0000 (11:26 +0100)]
mlxsw: spectrum_router: Flush FIB tables during fini
Since commit b45f64d16d45 ("mlxsw: spectrum_router: Use FIB notifications
instead of switchdev calls") we reflect to the device the entire FIB
table and not only FIBs that point to netdevs created by the driver.
During module removal, FIBs of the second type are removed following
NETDEV_UNREGISTER events sent. The other FIBs are still present in both
the driver's cache and the device's table.
Fix this by iterating over all the FIB tables in the device and flush
them. There's no need to take locks, as we're the only writer.
Fixes: b45f64d16d45 ("mlxsw: spectrum_router: Use FIB notifications instead of switchdev calls") Signed-off-by: Ido Schimmel <idosch@mellanox.com> Signed-off-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Mon, 14 Nov 2016 01:50:35 +0000 (17:50 -0800)]
net: stmmac: Fix lack of link transition for fixed PHYs
Commit 52f95bbfcf72 ("stmmac: fix adjust link call in case of a switch
is attached") added some logic to avoid polling the fixed PHY and
therefore invoking the adjust_link callback more than once, since this
is a fixed PHY and link events won't be generated.
This works fine the first time, because we start with phydev->irq =
PHY_POLL, so we call adjust_link, then we set phydev->irq =
PHY_IGNORE_INTERRUPT and we stop polling the PHY.
Now, if we called ndo_close(), which calls both phy_stop() and does an
explicit netif_carrier_off(), we end up with a link down. Upon calling
ndo_open() again, despite starting the PHY state machine, we have
PHY_IGNORE_INTERRUPT set, and we generate no link event at all, so the
link is permanently down.
Fixes: 52f95bbfcf72 ("stmmac: fix adjust link call in case of a switch is attached") Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Xin Long [Sun, 13 Nov 2016 13:44:37 +0000 (21:44 +0800)]
sctp: change sk state only when it has assocs in sctp_shutdown
Now when users shutdown a sock with SEND_SHUTDOWN in sctp, even if
this sock has no connection (assoc), sk state would be changed to
SCTP_SS_CLOSING, which is not as we expect.
Besides, after that if users try to listen on this sock, kernel
could even panic when it dereference sctp_sk(sk)->bind_hash in
sctp_inet_listen, as bind_hash is null when sock has no assoc.
This patch is to move sk state change after checking sk assocs
is not empty, and also merge these two if() conditions and reduce
indent level.
Fixes: d46e416c11c8 ("sctp: sctp should change socket state when shutdown is received") Reported-by: Andrey Konovalov <andreyknvl@google.com> Tested-by: Andrey Konovalov <andreyknvl@google.com> Signed-off-by: Xin Long <lucien.xin@gmail.com> Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 14 Nov 2016 21:20:54 +0000 (16:20 -0500)]
Merge branch 'bnx2-kdump-fix'
Baoquan He says:
====================
bnx2: Wait for in-flight DMA to complete at probe stage
This is v2 post.
In commit 3e1be7a ("bnx2: Reset device during driver initialization"),
firmware requesting code was moved from open stage to probe stage.
The reason is in kdump kernel hardware iommu need device be reset in
driver probe stage, otherwise those in-flight DMA from 1st kernel
will continue going and look up into the newly created io-page tables.
However bnx2 chip resetting involves firmware requesting issue, that
need be done in open stage.
Michale Chan suggested we can just wait for the old in-flight DMA to
complete at probe stage, then though without device resetting, we
don't need to worry the old in-flight DMA could continue looking up
the newly created io-page tables.
v1->v2:
Michael suggested to wait for the in-flight DMA to complete at probe
stage. So give up the old method of trying to reset chip at probe
stage, take the new way accordingly.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Baoquan He [Sun, 13 Nov 2016 05:01:33 +0000 (13:01 +0800)]
bnx2: Wait for in-flight DMA to complete at probe stage
In-flight DMA from 1st kernel could continue going in kdump kernel.
New io-page table has been created before bnx2 does reset at open stage.
We have to wait for the in-flight DMA to complete to avoid it look up
into the newly created io-page table at probe stage.
Suggested-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: Baoquan He <bhe@redhat.com> Acked-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net>
When people build bnx2 driver into kernel, it will fail to detect
and load firmware because firmware is contained in initramfs and
initramfs has not been uncompressed yet during do_initcalls. So
revert commit 3e1be7a and work out a new way in the later patch.
Signed-off-by: Baoquan He <bhe@redhat.com> Acked-by: Paul Menzel <pmenzel@molgen.mpg.de> Signed-off-by: David S. Miller <davem@davemloft.net>
Philippe Reynes [Sun, 13 Nov 2016 17:53:28 +0000 (18:53 +0100)]
net: atheros: atl2: use new api ethtool_{get|set}_link_ksettings
The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
The previous implementation of set_settings was modifying
the value of advertising, but with the new API, it's not
possible. The structure ethtool_link_ksettings is defined
as const.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Philippe Reynes [Sun, 13 Nov 2016 17:35:14 +0000 (18:35 +0100)]
net: atheros: atl1: use new api ethtool_{get|set}_link_ksettings
The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
The previous implementation of set_settings was modifying
the value of advertising, but with the new API, it's not
possible. The structure ethtool_link_ksettings is defined
as const.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
WANG Cong [Fri, 11 Nov 2016 18:20:50 +0000 (10:20 -0800)]
net: fix sleeping for sk_wait_event()
Similar to commit 14135f30e33c ("inet: fix sleeping inside inet_wait_for_connect()"),
sk_wait_event() needs to fix too, because release_sock() is blocking,
it changes the process state back to running after sleep, which breaks
the previous prepare_to_wait().
Switch to the new wait API.
Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Linus Torvalds [Mon, 14 Nov 2016 17:46:08 +0000 (09:46 -0800)]
ASoC: lpass-platform: fix uninitialized variable
In commit 022d00ee0b55 ("ASoC: lpass-platform: Fix broken pcm data
usage") the stream specific information initialization was broken, with
the dma channel information not being initialized if there was no
alloc_dma_channel() helper function.
Before that, the DMA channel number was implicitly initialized to zero
because the backing store was allocated with devm_kzalloc(). When the
init code was rewritten, that implicit initialization was lost, and gcc
rightfully complains about an uninitialized variable being used.
Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It turns out that this flushes things much too aggressiverly, and causes
lines to break up when the system logger races with new continuation
lines being printed.
There's a pending patch to make printk() flushing much more
straightforward, but it's too invasive for 4.9, so in the meantime let's
just not make the system message logging flush continuation lines.
They'll be flushed by the final newline anyway.
Suggested-by: Petr Mladek <pmladek@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This file was converted to a separate module at commit 7a0786c19d65
("gp8psk: Fix DVB frontend attach"), because the DVB attach routines
require it to work. However, I forgot to copy the MODULE_foo() macros
from the original module, causing this warning:
WARNING: modpost: missing MODULE_LICENSE() in drivers/media/dvb-frontends/gp8psk-fe.o
projects / mirror_ubuntu-focal-kernel.git / log