Extend ndo_setup_tc() to support ingress tc offloading. Will be used by
later patches to offload tc flower filter.
Feature is off by default and could be enabled by issuing:
# ethtool -K eth0 hw-tc-offload on
Offloads flow table is dynamically created when first filter is
added.
Rules are saved in a hash table that is maintained by the consumer (for
example - the flower offload in the next patch).
When last filter is removed and no filters exist in the hash table, the
offload flow table is destroyed.
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the vlan and main flow tables to use priority 1. This will allow
the upcoming TC offload logic to use a higher priority (0) for the
offload steering table.
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Restricting handle to TC_H_ROOT breaks the old instantiation of mqprio
to setup a hardware qdisc. This patch relaxes the test, to only check the
type.
Fixes: 08fb1da ("net/mlx5e: Support DCBNL IEEE ETS")
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to handle flow table entry destinations only if the action
associated with the rule is forwarding (MLX5_FLOW_CONTEXT_ACTION_FWD_DEST).
Fixes: 26a8145390 ('net/mlx5_core: Introduce flow steering firmware commands')
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Enable device drivers to query the action, if and only if is a mark
action and what value to use for marking.
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce the macros tc_no_actions and tc_for_each_action to make code
clearer.
Extracted struct tc_action out of the ifdef to make calls to
is_tcf_gact_shot() and similar functions valid, even when it is a nop.
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Suggested-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
Will be used in a following patch to query if a key is being used, and
what it's value in the target object.
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is based on a patch made by John Fastabend.
It adds support for offloading cls_flower.
when NETIF_F_HW_TC is on:
flags = 0 => Rule will be processed twice - by hardware, and if
still relevant, by software.
flags = SKIP_HW => Rull will be processed by software only
If hardware fail/not capabale to apply the rule, operation will NOT
fail. Filter will be processed by SW only.
Acked-by: Jiri Pirko <jiri@mellanox.com>
Suggested-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
John Crispin says:
====================
net-next: mediatek: add ethernet driver
This series adds support for the Mediatek ethernet core found on current ARM
based SoCs. The driver works on MT2701 and MT7623 SoCs
Instead of trying to upstream everything at once I decided to concentrate on
the important parts required to make current generation silicon work. The V3
series only includes the code required to make dual MAC setups work and only
supports the newer QDMA engine.
Changes in V5
* reduce the mdio timeut to HZ
* add a call to usleep_range() which schedules in the background.
Changes in V4
* remove ugly _FE macro, use offsetof() instead
Changes in V3
* only include code for MT2701/7623 support
* drop support for PDMA and older MIPS based SoCs
* drop switch support
Changes in V2
* change the namespace of the functions from fe_* to mtk_*
* add support for the latest generation of ARM SoCs
* add dual MAC support
* remove the swconfig specific bits
* remove most of the magic values and replace them with defines
* add verbose descriptions to the patches
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add myself and Felix as the Maintainers for the MediaTek ethernet driver.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the Makefile and Kconfig required to make the driver build.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add ethernet support for MediaTek SoCs from the MT7623 family. These have
dual GMAC. Depending on the exact version, there might be a built-in
Gigabit switch (MT7530). The core does not have the typical DMA ring setup.
Instead there is a linked list that we add descriptors to. There is only
one linked list that both MACs use together. There is a special field
inside the TX descriptors called the VQID. This allows us to assign packets
to different internal queues. By using a separate id for each MAC we are
able to get deterministic results for BQL. Additionally we need to
provide the core with a block of scratch memory that is the same size as
the RX ring and data buffer. This is really needed to make the HW datapath
work. Although the driver does not support this yet, we still need to
assign the memory and tell the core about it for RX to work.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Michael Lee <igvtee@gmail.com>
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds the binding documentation for the MediaTek Ethernet
controller.
Signed-off-by: John Crispin <blogic@openwrt.org>
Acked-by: Rob Herring <robh@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
The initial commit badly merged into the dsa_resume method instead
of the dsa_remove_dst method.
As consequence, the dst->master_netdev->dsa_ptr is not set to NULL on
removal and re-bind of the dsa device fails with error -17.
Fixes: b0dc635d92 ("net: dsa: cleanup resources upon module removal ")
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
'commit 55482edc25
("qede: Add slowpath/fastpath support and enable hardware GRO")'
introduces below error when compiling net-next with "make ARCH=x86_64"
drivers/built-in.o: In function `qede_rx_int':
qede_main.c:(.text+0x6101a0): undefined reference to `tcp_gro_complete'
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rajesh Borundia says:
====================
qlcnic fixes
This series adds following fixes.
o While processing mailbox if driver gets a spurious mailbox
interrupt it leads into premature completion of a next
mailbox request. Added a guard against this by checking current
state of mailbox and ignored spurious interrupt.
Added a stats counter to record this condition.
v2:
o Added patch that removes usage of atomic_t as we are not implemeting
atomicity by using atomic_t value.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
o While the driver is in the middle of a MB completion processing
and it receives a spurious MB interrupt, it is mistaken as a good MB
completion interrupt leading to premature completion of the next MB
request. Fix the driver to guard against this by checking the current
state of MB processing and ignore the spurious interrupt.
Also added a stats counter to record this condition.
Signed-off-by: Rajesh Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
o atomic_t usage is incorrect as we are not implementing
any atomicity.
Signed-off-by: Rajesh Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hariprasad Shenai says:
====================
cxgb4vf: Interrupt and queue configuration changes
This series fixes some issues and some changes in the queue and interrupt
configuration for cxgb4vf driver. We need to enable interrupts before we
register our network device, so that we don't loose link up interrupts.
Allocate rx queues based on interrupt type. Set number of tx/rx queues in
probe function only. Also adds check for some invalid configurations.
This patch series has been created against net-next tree and includes
patches on cxgb4vf driver.
We have included all the maintainers of respective drivers. Kindly review
the change and let us know in case of any review comments.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The Queue Set Configuration code was always reserving room for a
Forwarded interrupt Queue even in the cases where we weren't using it.
Figure out how many Ports and Queue Sets we can support. This depends on
knowing our Virtual Function Resources and may be called a second time
if we fall back from MSI-X to MSI Interrupt Mode. This change fixes that
problem.
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This avoids a race condition where a system that has network devices set up
to be automatically configured and we get the first Port Link Status
message from the firmware on the Asynchronous Firmware Event Queue before
we've enabled interrupts. If that happens, we end up losing the interrupt
and never realizing that the links has actually come up.
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no need to change the 802.1Q port mode for the same value.
Thus avoid such message:
[ 401.954836] dsa dsa@0 lan0: 802.1Q Mode: Disabled (was Disabled)
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The port register 0x07 contains more options than just the default VID,
even though they are not used yet. So prefer a read then write operation
over a direct write.
This also allows to keep track of the change through dynamic debug.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Apply a few non-functional changes on the port state setter:
* add a dynamic debug message with state names to track changes
* explicit states checking instead of assuming their numeric values
* lock mutex only once when changing several port states
* use bitmap macros to declare and access port_state_update_mask
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sergei Shtylyov says:
====================
sh_eth: fix couple of bugs in sh_eth_ring_format()
Here's a set of 2 patches against DaveM's 'net.git' repo fixing two bugs
in sh_eth_.ring_format()...
[1/2] sh_eth: fix NULL pointer dereference in sh_eth_ring_format()
[2/2] sh_eth: advance 'rxdesc' later in sh_eth_ring_format()
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Iff dma_map_single() fails, 'rxdesc' should point to the last filled RX
descriptor, so that it can be marked as the last one, however the driver
would have already advanced it by that time. In order to fix that, only
fill an RX descriptor once all the data for it is ready.
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a low memory situation, if netdev_alloc_skb() fails on a first RX ring
loop iteration in sh_eth_ring_format(), 'rxdesc' is still NULL. Avoid
kernel oops by adding the 'rxdesc' check after the loop.
Reported-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The stub helper functions for the newly added kcm_proc_init/exit interfaces
are defined as 'static' in a header file, which leads to build warnings for
each file that includes them without calling them:
include/net/kcm.h:183:12: error: 'kcm_proc_init' defined but not used [-Werror=unused-function]
include/net/kcm.h:184:13: error: 'kcm_proc_exit' defined but not used [-Werror=unused-function]
This marks the two functions as 'static inline' instead, which avoids the
warnings and is obviously what was meant here.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: cd6e111bf5 ("kcm: Add statistics and proc interfaces")
Signed-off-by: David S. Miller <davem@davemloft.net>
-----BEGIN PGP SIGNATURE-----
iQEcBAABCgAGBQJW4T63AAoJED07qiWsqSVq36MIAIqx00Qqk73/erXp1kWGx4QH
3jv26HzjmEP6fxBFeduNMMT/jgcNhXkrDru8LhJPGcoOqftdJOaHB38J+DEXYjNZ
jwnKEc4rVPd4o8z6ApoXHS72nucYus4L/5JSWx5nmiNJ3pW1j2Xnb3HfcInt5XyK
pwZPedfriKmPxmSH0Gvn82imyGaiJ/63GNYse2BwsdZZrff3nTG+nFFj2cHpeTSR
oQEvXQUBtjk/4lPCyR0rf1C8ZMMxh/L8ovDTFTE0p9asupcEEbc8LIPLy8ZqM8Zo
KzC/3eCbgCgPiwqqgBLUJXHjTviwb/1AHfZ4KtnpLaqyuyomSAXaUEITAp2n5Ns=
=+E/w
-----END PGP SIGNATURE-----
Merge tag 'linux-can-next-for-4.6-20160310' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next
Marc Kleine-Budde says:
====================
pull-request: can-next 2016-03-10
this is a pull request of 5 patch for net-next/master.
Marek Vasut contributes 4 patches for the ifi CAN driver, which makes
it work on real hardware. There is one patch by Ramesh Shanmugasundaram
for the rcar_can driver that adds support for the 3rd generation IP
core.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of CAN2.0 EFF frame, the controller handles frame IDs in a
rather bizzare way. The ID is split into an extended part, IDX[28:11]
and standard part, ID[10:0]. In the TX path, the core first sends the
top 11 bits of the IDX, followed by ID and finally the rest of IDX.
In the RX path, the core stores the ID the LSbit part of IDX field,
followed by the LSbit parts of real IDX. The MSbit parts of IDX are
stored in ID field of the register.
This patch implements the necessary bit shuffling to mitigate this
obscure behavior. In case two of these controllers are connected
together, the RX and TX bit swapping nullifies itself and the issue
does not manifest. The issue only manifests when talking to another
different CAN controller.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The RX and TX ID mask for CAN2.0 is 11 bits wide. This patch fixes
the incorrect mask, which caused the CAN IDs to miss the MSBit both
on receive and transmit.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The TX DLC, the transmission length information, was not written
into the transmit configuration register. When using the CAN core
with different CAN controller, the receiving CAN controller will
receive only the ID part of the CAN frame, but no data at all.
This patch adds the TX DLC into the register to fix this issue.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The clock generation does not match reality when using the CAN IP
core outside of the FPGA design. This patch fixes the computation
of values which are programmed into the clock generator registers.
First, there are some off-by-one errors which manifest themselves
only when communicating with different controller, so those are
fixed.
Second, the bits in the clock generator registers have different
meaning depending on whether the core is in ISO CANFD mode or any
of the other modes (BOSCH CANFD or CAN2.0). Detect the ISO CANFD
mode and fix handling of this special case of clock configuration.
Finally, the CAN clock speed is in CANCLOCK register, not SYSCLOCK
register, so fix this as well.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Lots of places in the kernel use memcpy(buf, comm, TASK_COMM_LEN); but
the result is typically passed to print("%s", buf) and extra bytes
after zero don't cause any harm.
In bpf the result of bpf_get_current_comm() is used as the part of
map key and was causing spurious hash map mismatches.
Use strlcpy() to guarantee zero-terminated string.
bpf verifier checks that output buffer is zero-initialized,
so even for short task names the output buffer don't have junk bytes.
Note it's not a security concern, since kprobe+bpf is root only.
Fixes: ffeedafbf0 ("bpf: introduce current->pid, tgid, uid, gid, comm accessors")
Reported-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
0-day bot reported build error:
kernel/built-in.o: In function `map_lookup_elem':
>> kernel/bpf/.tmp_syscall.o:(.text+0x329b3c): undefined reference to `bpf_stackmap_copy'
when CONFIG_BPF_SYSCALL is set and CONFIG_PERF_EVENTS is not.
Add weak definition to resolve it.
This code path in map_lookup_elem() is never taken
when CONFIG_PERF_EVENTS is not set.
Fixes: 557c0c6e7d ("bpf: convert stackmap to pre-allocation")
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Willem de Bruijn says:
====================
net: validate variable length ll headers
Allow device-specific validation of link layer headers. Existing
checks drop all packets shorter than hard_header_len. For variable
length protocols, such packets can be valid.
patch 1 adds header_ops.validate and dev_validate_header
patch 2 implements the protocol specific callback for AX25
patch 3 replaces ll_header_truncated with dev_validate_header
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace link layer header validation check ll_header_truncate with
more generic dev_validate_header.
Validation based on hard_header_len incorrectly drops valid packets
in variable length protocols, such as AX25. dev_validate_header
calls header_ops.validate for such protocols to ensure correctness
below hard_header_len.
See also http://comments.gmane.org/gmane.linux.network/401064
Fixes 9c7077622d ("packet: make packet_snd fail on len smaller than l2 header")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As variable length protocol, AX25 fails link layer header validation
tests based on a minimum length. header_ops.validate allows protocols
to validate headers that are shorter than hard_header_len. Implement
this callback for AX25.
See also http://comments.gmane.org/gmane.linux.network/401064
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Netdevice parameter hard_header_len is variously interpreted both as
an upper and lower bound on link layer header length. The field is
used as upper bound when reserving room at allocation, as lower bound
when validating user input in PF_PACKET.
Clarify the definition to be maximum header length. For validation
of untrusted headers, add an optional validate member to header_ops.
Allow bypassing of validation by passing CAP_SYS_RAWIO, for instance
for deliberate testing of corrupt input. In this case, pad trailing
bytes, as some device drivers expect completely initialized headers.
See also http://comments.gmane.org/gmane.linux.network/401064
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tom Herbert says:
====================
kcm: Kernel Connection Multiplexor (KCM)
Kernel Connection Multiplexor (KCM) is a facility that provides a
message based interface over TCP for generic application protocols.
The motivation for this is based on the observation that although
TCP is byte stream transport protocol with no concept of message
boundaries, a common use case is to implement a framed application
layer protocol running over TCP. To date, most TCP stacks offer
byte stream API for applications, which places the burden of message
delineation, message I/O operation atomicity, and load balancing
in the application. With KCM an application can efficiently send
and receive application protocol messages over TCP using a
datagram interface.
In order to delineate message in a TCP stream for receive in KCM, the
kernel implements a message parser. For this we chose to employ BPF
which is applied to the TCP stream. BPF code parses application layer
messages and returns a message length. Nearly all binary application
protocols are parsable in this manner, so KCM should be applicable
across a wide range of applications. Other than message length
determination in receive, KCM does not require any other application
specific awareness. KCM does not implement any other application
protocol semantics-- these are are provided in userspace or could be
implemented in a kernel module layered above KCM.
KCM implements an NxM multiplexor in the kernel as diagrammed below:
+------------+ +------------+ +------------+ +------------+
| KCM socket | | KCM socket | | KCM socket | | KCM socket |
+------------+ +------------+ +------------+ +------------+
| | | |
+-----------+ | | +----------+
| | | |
+----------------------------------+
| Multiplexor |
+----------------------------------+
| | | | |
+---------+ | | | ------------+
| | | | |
+----------+ +----------+ +----------+ +----------+ +----------+
| Psock | | Psock | | Psock | | Psock | | Psock |
+----------+ +----------+ +----------+ +----------+ +----------+
| | | | |
+----------+ +----------+ +----------+ +----------+ +----------+
| TCP sock | | TCP sock | | TCP sock | | TCP sock | | TCP sock |
+----------+ +----------+ +----------+ +----------+ +----------+
The KCM sockets provide the datagram interface to applications,
Psocks are the state for each attached TCP connection (i.e. where
message delineation is performed on receive).
A description of the APIs and design can be found in the included
Documentation/networking/kcm.txt.
In this patch set:
- Add MSG_BATCH flag. This is used in sendmsg msg_hdr flags to
indicate that more messages will be sent on the socket. The stack
may batch messages up if it is beneficial for transmission.
- In sendmmsg, set MSG_BATCH in all sub messages except for the last
one.
- In order to allow sendmmsg to contain multiple messages with
SOCK_SEQPAKET we allow each msg_hdr in the sendmmsg to set MSG_EOR.
- Add KCM module
- This supports SOCK_DGRAM and SOCK_SEQPACKET.
- KCM documentation
v2:
- Added splice and page operations.
- Assemble receive messages in place on TCP socket (don't have a
separate assembly queue.
- Based on above, enforce maxmimum receive message to be the size
of the recceive socket buffer.
- Support message assembly timeout. Use the timeout value in
sk_rcvtimeo on the TCP socket.
- Tested some with a couple of other production applications,
see ~5% improvement in application latency.
Testing:
Dave Watson has integrated KCM into Thrift and we intend to put these
changes into open source. Example of this is in:
https://github.com/djwatson/fbthrift/commit/
dd7e0f9cf4e80912fdb90f6cd394db24e61a14cc
Some initial KCM Thrift benchmark numbers (comment from Dave)
Thrift by default ties a single connection to a single thread. KCM is
instead able to load balance multiple connections across multiple epoll
loops easily.
A test sending ~5k bytes of data to a kcm thrift server, dropping the
bytes on recv:
QPS Latency / std dev Latency
without KCM
70336 209/123
with KCM
70353 191/124
A test sending a small request, then doing work in the epoll thread,
before serving more requests:
QPS Latency / std dev Latency
without KCM
14282 559/602
with KCM
23192 344/234
At the high end, there's definitely some additional kernel overhead:
Cranking the pipelining way up, with lots of small requests
QPS Latency / std dev Latency
without KCM
1863429 127/119
with KCM
1337713 192/241
---
So for a "realistic" workload, KCM performs pretty well (second case).
Under extreme conditions of highest tps we still have some work to do.
In its nature a multiplexor will spread work between CPUs which is
logically good for load balancing but coan conflict with the goal
promoting affinity. Batching messages on both send and receive are
the means to recoup performance.
Future support:
- Integration with TLS (TLS-in-kernel is a separate initiative).
- Page operations/splice support
- Unconnected KCM sockets. Will be able to attach sockets to different
destinations, AF_KCM addresses with be used in sendmsg and recvmsg
to indicate destination
- Explore more utility in performing BPF inline with a TCP data stream
(setting SO_MARK, rxhash for messages being sent received on
KCM sockets).
- Performance work
- Diagnose performance issues under high message load
FAQ (Questions posted on LWN)
Q: Why do this in the kernel?
A: Because the kernel is good at scheduling threads and steering packets
to threads. KCM fits well into this model since it allows the unit
of work for scheduling and steering to be the application layer
messages themselves. KCM should be thought of as generic application
protocol acceleration. It to the philosophy that the kernel provides
generic and extensible interfaces.
Q: How can adding code in the path yield better performance?
A: It is true that for just sending receiving a single message there
would be some performance loss since the code path is longer (for
instance comparing netperf to KCM). But for real production
applications performance takes on many dynamics. Parallelism, context
switching, affinity, granularity of locking, and load balancing are
all relevant. The theory of KCM is that by an application-centric
interface, the kernel can provide better support for these
performance characteristics.
Q: Why not use an existing message-oriented protocol such as RUDP,
DCCP, SCTP, RDS, and others?
A: Because that would entail using a completely new transport protocol.
Deploying a new protocol at scale is either a huge undertaking or
fundamentally infeasible. This is true in either the Internet and in
the data center due in a large part to protocol ossification.
Besides, KCM we want KCM to work existing, well deployed application
protocols that we couldn't change even if we wanted to (e.g. http/2).
KCM simply defines a new interface method, it does not redefine any
aspect of the transport protocol nor application protocol, nor set
any new requirements on these. Neither does KCM attempt to implement
any application protocol logic other than message deliniation in the
stream. These are fundamental requirement of KCM.
Q: How does this affect TCP?
A: It doesn't, not in the slightest. The use of KCM can be one-sided,
KCM has no effect on the wire.
Q: Why force TCP into doing something it's not designed for?
A: TCP is defined as transport protocol and there is no standard that
says the API into TCP must be stream based sockets, or for that
matter sockets at all (or even that TCP needs to be implemented in a
kernel). KCM is not inconsistent with the design of TCP just because
to makes an message based interface over TCP, if it were then every
application protocol sending messages over TCP would also be! :-)
Q: What about the problem of a connections with very slow rate of
incoming data? As a result your application can get storms of very
short reads. And it actually happens a lot with connection from
mobile devices and it is a problem for servers handling a lot of
connections.
A: The storm of short reads will occur regardless of whether KCM is used
or not. KCM does have one advantage in this scenario though, it will
only wake up the application when a full message has been received,
not for each packet that makes up part of a bigger messages. If a
bunch of small messages are received, the application can receive
messages in batches using recvmmsg.
Q: Why not just use DPDK, or at least provide KCM like functionality in
DPDK?
A: DPDK, or more generally OS bypass presumably with a TCP stack in
userland, presents a different model of load balancing than that of
KCM (and the kernel). KCM implements load balancing of messages
across the threads of an application, whereas DPDK load balances
based on queues which are more static and coarse-grained since
multiple connections are bound to queues. DPDK works best when
processing of packets is silo'ed in a thread on the CPU processing
a queue, and packet processing (for both the stack and application)
is fairly uniform. KCM works well for applications where the amount
of work to process messages varies an application work is commonly
delegated to worker threads often on different CPUs.
The message based interface over TCP is something that could be
provide by a DPDK or OS bypass library.
Q: I'm not quite seeing this for HTTP. Maybe for HTTP/2, I guess, or web
sockets?
A: Yes. KCM is most appropriate for message based protocols over TCP
where is easy to deduce the message length (e.g. a length field)
and the protocol implements its own message ordering semantics.
Fortunately this encompasses many modern protocols.
Q: How is memory limited and controlled?
A: In v2 all data for messages is now kept in socket buffers, either
those for TCP or KCM, so socket buffer limits are applicable.
This includes receive messages assembly which is now done ont teh
TCP socket buffer instead of a separate queue-- this has the
consequence that the TCP socket buffer limit provides an
enforceable maxmimum message size.
Additionally, a timeout may be set for messages assembly. The
value used for this is taken from sk_rcvtimeo of the TCP socket.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds receive timeout for message assembly on the attached TCP
sockets. The timeout is set when a new messages is started and the whole
message has not been received by TCP (not in the receive queue). If the
completely message is subsequently received the timer is cancelled, if the
timer expires the RX side is aborted.
The timeout value is taken from the socket timeout (SO_RCVTIMEO) that is
set on a TCP socket (i.e. set by get sockopt before attaching a TCP socket
to KCM.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Message assembly is performed on the TCP socket. This is logically
equivalent of an application that performs a peek on the socket to find
out how much memory is needed for a receive buffer. The receive socket
buffer also provides the maximum message size which is checked.
The receive algorithm is something like:
1) Receive the first skbuf for a message (or skbufs if multiple are
needed to determine message length).
2) Check the message length against the number of bytes in the TCP
receive queue (tcp_inq()).
- If all the bytes of the message are in the queue (incluing the
skbuf received), then proceed with message assembly (it should
complete with the tcp_read_sock)
- Else, mark the psock with the number of bytes needed to
complete the message.
3) In TCP data ready function, if the psock indicates that we are
waiting for the rest of the bytes of a messages, check the number
of queued bytes against that.
- If there are still not enough bytes for the message, just
return
- Else, clear the waiting bytes and proceed to receive the
skbufs. The message should now be received in one
tcp_read_sock
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement kcm_sendpage. Set in sendpage to kcm_sendpage in both
dgram and seqpacket ops.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement kcm_splice_read. This is supported only for seqpacket.
Add kcm_seqpacket_ops and set splice read to kcm_splice_read.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds various counters for KCM. These include counters for
messages and bytes received or sent, as well as counters for number of
attached/unattached TCP sockets and other error or edge events.
The statistics are exposed via a proc interface. /proc/net/kcm provides
statistics per KCM socket and per psock (attached TCP sockets).
/proc/net/kcm_stats provides aggregate statistics.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This module implements the Kernel Connection Multiplexor.
Kernel Connection Multiplexor (KCM) is a facility that provides a
message based interface over TCP for generic application protocols.
With KCM an application can efficiently send and receive application
protocol messages over TCP using datagram sockets.
For more information see the included Documentation/networking/kcm.txt
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Amir Vadai
David S. Miller
John Crispin
Neil Armstrong
Manish Chopra
Rajesh Borundia
Hariprasad Shenai
Vivien Didelot
Sergei Shtylyov
Arnd Bergmann
Ramesh Shanmugasundaram
Marek Vasut
Alexei Starovoitov
Willem de Bruijn
Tom Herbert