The DongShanPiOne is little SigmaStar SSD202D based module
from 100ask.
Add an initial dts for this board.
Link: http://linux-chenxing.org/infinity2/dongshanpione/
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
The DongShanPiOne is a SigmaStar SSD202D based module.
Add a compatible for it to the list of infinity2m boards.
Link: http://linux-chenxing.org/infinity2/dongshanpione/
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Acked-by: Rob Herring <robh@kernel.org>
100ask is Linux SBC/module/training provider.
Add a prefix for them.
Link: https://www.100ask.net
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Acked-by: Rob Herring <robh@kernel.org>
Miyoo has released a portable retro games machine based
on the SigmaStar SSD202D. Add the initial device tree
for this machine. Just enough to get to an initramfs
shell.
Link: http://linux-chenxing.org/infinity2/miyoomini/
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
The Miyoo Mini is a SigmaStar SSD202D based retro emulation
device.
Add a compatible for it to the list of infinity2m devices.
Link: http://linux-chenxing.org/infinity2/miyoomini/
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Acked-by: Rob Herring <robh@kernel.org>
Add a prefix for "miyoo". The only details I can find about
the company is their aliexpress store "miyoo global store":
https://www.aliexpress.com/store/912663639
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Acked-by: Rob Herring <robh@kernel.org>
The Wireless Tag IDO-SBC2D06-V1B-22W[1] is an SBC powered by SSD202D
with a dual Ethernet and a connector for a 4-inch and 7-inch display.
It embeds a System-On-Module IDO-SOM2D01[2] with an Mstar SSD202 SoC
dual-core Cortex-A7 CPU @ 1.2Ghz , 2D GPU, 128 MB DDR3 (on-chip) and
a 256MB SPI NAND flash. This commit adds basic definition for this
board.
1. http://linux-chenxing.org/infinity2/ido-sbc2d06
2. http://www.wireless-tag.com/portfolio/ido-som2d01
Signed-off-by: Romain Perier <romain.perier@gmail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
This adds a vendor prefix for wireless tag boards and SOMs.
Signed-off-by: Romain Perier <romain.perier@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Now there is gpio support for ssd20xd set the right compatible in the gpio node.
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
Reviewed-by: Romain Perier <romain.perier@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
We are currently computing the relative data rates as
src_size * scale_factor where scale_factor is src_size / dst_size.
Thus relative data rate is src_size * src_size / dst_size,
which is just utter nonsense. What we really seem to want is
just a reasonable estimate on how much data will be fetched
which is just src_size. So let's do that instead.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220118092354.11631-4-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
The kernel test robot is reporting that xfstest which does
umount ext2 on xfs
umount xfs
sequence started failing, for commit 322c4293ec ("loop: make
autoclear operation asynchronous") removed a guarantee that fput() of
backing file is processed before lo_release() from close() returns to
user mode.
And syzbot is reporting that deferring destroy_workqueue() from
__loop_clr_fd() to a WQ context did not help [1]. Revert that commit.
Link: https://syzkaller.appspot.com/bug?extid=831661966588c802aae9 [1]
Reported-by: kernel test robot <oliver.sang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reported-by: syzbot <syzbot+831661966588c802aae9@syzkaller.appspotmail.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Link: https://lore.kernel.org/r/20220211071554.3424-1-penguin-kernel@I-love.SAKURA.ne.jp
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If bpf_msg_push_data() is called with len 0 (as it happens during
selftests/bpf/test_sockmap), we do not need to do anything and can
return early.
Calling bpf_msg_push_data() with len 0 previously lead to a wrong ENOMEM
error: we later called get_order(copy + len); if len was 0, copy + len
was also often 0 and get_order() returned some undefined value (at the
moment 52). alloc_pages() caught that and failed, but then bpf_msg_push_data()
returned ENOMEM. This was wrong because we are most probably not out of
memory and actually do not need any additional memory.
Fixes: 6fff607e2f ("bpf: sk_msg program helper bpf_msg_push_data")
Signed-off-by: Felix Maurer <fmaurer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/df69012695c7094ccb1943ca02b4920db3537466.1644421921.git.fmaurer@redhat.com
This patch adds the device tree to support Toradex Verdin iMX8M Mini a
computer on module which can be used on different carrier boards.
The module consists of an NXP i.MX 8M Mini family SoC (either i.MX 8M
Mini Quad or 8M Mini DualLite), a PCA9450A PMIC, a Gigabit Ethernet PHY,
1 or 2 GB of LPDDR4 RAM, an eMMC, a TLA2024 ADC, an I2C EEPROM, an
RX8130 RTC, an optional SPI CAN controller plus an optional Bluetooth/
Wi-Fi module.
Anything that is not self-contained on the module is disabled by
default.
The device tree for the Dahlia includes the module's device tree and
enables the supported peripherals of the carrier board.
The device tree for the Verdin Development Board includes the module's
device tree as well as the Dahlia one as it is a superset and supports
almost all peripherals available.
So far there is no display functionality supported at all but basic
console UART, PCIe, USB host, eMMC and Ethernet and PCIe functionality
work fine.
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Add toradex,verdin-imx8mm for our new Verdin iMX8M Mini modules, its
nonwifi and wifi variants and the carrier boards (both Dahlia and the
Verdin Development Board) they may be mated in.
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Add interrupt controller mode for the pca6416 on i.MX8MP EVK board's.
Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
With commit 15d16d6dad ("kbuild: Add generic rule to apply
fdtoverlay"), overlay target can be used to simplify the build of DTB
overlays. It also performs a cross check to ensure base DT and overlay
actually match.
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
As suggested by commit 9ae8578b51 ("of: Documentation: change overlay
example to use current syntax"), there is no need to have overlay syntax
be hard coded in the device tree source file any more.
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains Netfilter fixes for net:
1) Add selftest for nft_synproxy, from Florian Westphal.
2) xt_socket destroy path incorrectly disables IPv4 defrag for
IPv6 traffic (typo), from Eric Dumazet.
3) Fix exit value selftest nft_concat_range.sh, from Hangbin Liu.
4) nft_synproxy disables the IPv4 hooks if the IPv6 hooks fail
to be registered.
5) disable rp_filter on router in selftest nft_fib.sh, also
from Hangbin Liu.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
ipv6: remove addrconf reliance on loopback
Second patch in this series removes IPv6 requirement about the netns
loopback device being the last device being dismantled.
This was needed because rt6_uncached_list_flush_dev()
and ip6_dst_ifdown() had to switch dst dev to a known
device (loopback).
Instead of loopback, we can use the (hidden) blackhole_netdev
which is also always there.
This will allow future simplfications of netdev_run_to()
and other parts of the stack like default_device_exit_batch().
Last two patches are optimizations for both IP families.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This is an optimization to keep the per-cpu lists as short as possible:
Whenever rt_flush_dev() changes one rtable dst.dev
matching the disappearing device, it can can transfer the object
to a quarantine list, waiting for a final rt_del_uncached_list().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is an optimization to keep the per-cpu lists as short as possible:
Whenever rt6_uncached_list_flush_dev() changes one rt6_info
matching the disappearing device, it can can transfer the object
to a quarantine list, waiting for a final rt6_uncached_list_del().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IPv6 addrconf notifiers wants the loopback device to
be the last device being dismantled at netns deletion.
This caused many limitations and work arounds.
Back in linux-5.3, Mahesh added a per host blackhole_netdev
that can be used whenever we need to make sure objects no longer
refer to a disappearing device.
If we attach to blackhole_netdev an ip6_ptr (allocate an idev),
then we can use this special device (which is never freed)
in place of the loopback_dev (which can be freed).
This will permit improvements in netdev_run_todo() and other parts
of the stack where had steps to make sure loopback_dev was
the last device to disappear.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This counter has never been visible, there is little point
trying to maintain it.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever Full soft reset is required, reset all individual engines
first, and then do a full soft reset.
Signed-off-by: Bruce Chang <yu.bruce.chang@intel.com>
cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220128185209.18077-5-ramalingam.c@intel.com
On i915_selftest@live@gt_timelines, we create many contexts in loop and
create and submit request and then destoy contexts. Destroying the context
needs to disable scheduling, wait for G2H, deregister context and wait
for G2H to destroy each context. Idling of the gt has to wait for all
this to complete which is taking ~3sec for this test.
Hence we are increasing the igt_flush_test's timeout for gt idling to
3Sec.
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220128185209.18077-4-ramalingam.c@intel.com
An indirect ctx wabb is implemented as per Wa_22011450934 to avoid rcs
restore hang during context restore of a preempted context in GPGPU mode
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
cc: Chris Wilson <chris.p.wilson@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220128185209.18077-2-ramalingam.c@intel.com
commit c50df701d4 ("drm/i915: Enable rpm wakeref tracking
whether runtime pm is enabled or not") enabled wakeref tracking
even for the mock device. Turns out that has somewhat significant
overhead, and on the glacial Core m3's we have in CI the vma
selftests are now exceeding the allotted time budget.
So let's disable the wakeref tracking once again for the mock
device in order to avoid blowing up the selftest runtime.
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Acked-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220204171053.18409-1-ville.syrjala@linux.intel.com
The mv88e6352, mv88e6240 and mv88e6176 have a serdes interface. This patch
allows to configure the output swing to a desired value in the
phy-handle of the port. The value which is peak to peak has to be
specified in microvolts. As the chips only supports eight dedicated
values we return EINVAL if the value in the DTS does not match one of
these values.
Signed-off-by: Holger Brunck <holger.brunck@hitachienergy.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Marek Behún <kabel@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Common PHYs and network PCSes often have the possibility to specify
peak-to-peak voltage on the differential pair - the default voltage
sometimes needs to be changed for a particular board.
Add properties `tx-p2p-microvolt` and `tx-p2p-microvolt-names` for this
purpose. The second property is needed to specify the mode for the
corresponding voltage in the `tx-p2p-microvolt` property, if the voltage
is to be used only for speficic mode. More voltage-mode pairs can be
specified.
Example usage with only one voltage (it will be used for all supported
PHY modes, the `tx-p2p-microvolt-names` property is not needed in this
case):
tx-p2p-microvolt = <915000>;
Example usage with voltages for multiple modes:
tx-p2p-microvolt = <915000>, <1100000>, <1200000>;
tx-p2p-microvolt-names = "2500base-x", "usb", "pcie";
Add these properties into a separate file phy/transmit-amplitude.yaml,
which should be referenced by any binding that uses it.
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
trace_napi_poll_hit() is reading stat->dev while another thread can write
on it from dropmon_net_event()
Use READ_ONCE()/WRITE_ONCE() here, RCU rules are properly enforced already,
we only have to take care of load/store tearing.
BUG: KCSAN: data-race in dropmon_net_event / trace_napi_poll_hit
write to 0xffff88816f3ab9c0 of 8 bytes by task 20260 on cpu 1:
dropmon_net_event+0xb8/0x2b0 net/core/drop_monitor.c:1579
notifier_call_chain kernel/notifier.c:84 [inline]
raw_notifier_call_chain+0x53/0xb0 kernel/notifier.c:392
call_netdevice_notifiers_info net/core/dev.c:1919 [inline]
call_netdevice_notifiers_extack net/core/dev.c:1931 [inline]
call_netdevice_notifiers net/core/dev.c:1945 [inline]
unregister_netdevice_many+0x867/0xfb0 net/core/dev.c:10415
ip_tunnel_delete_nets+0x24a/0x280 net/ipv4/ip_tunnel.c:1123
vti_exit_batch_net+0x2a/0x30 net/ipv4/ip_vti.c:515
ops_exit_list net/core/net_namespace.c:173 [inline]
cleanup_net+0x4dc/0x8d0 net/core/net_namespace.c:597
process_one_work+0x3f6/0x960 kernel/workqueue.c:2307
worker_thread+0x616/0xa70 kernel/workqueue.c:2454
kthread+0x1bf/0x1e0 kernel/kthread.c:377
ret_from_fork+0x1f/0x30
read to 0xffff88816f3ab9c0 of 8 bytes by interrupt on cpu 0:
trace_napi_poll_hit+0x89/0x1c0 net/core/drop_monitor.c:292
trace_napi_poll include/trace/events/napi.h:14 [inline]
__napi_poll+0x36b/0x3f0 net/core/dev.c:6366
napi_poll net/core/dev.c:6432 [inline]
net_rx_action+0x29e/0x650 net/core/dev.c:6519
__do_softirq+0x158/0x2de kernel/softirq.c:558
do_softirq+0xb1/0xf0 kernel/softirq.c:459
__local_bh_enable_ip+0x68/0x70 kernel/softirq.c:383
__raw_spin_unlock_bh include/linux/spinlock_api_smp.h:167 [inline]
_raw_spin_unlock_bh+0x33/0x40 kernel/locking/spinlock.c:210
spin_unlock_bh include/linux/spinlock.h:394 [inline]
ptr_ring_consume_bh include/linux/ptr_ring.h:367 [inline]
wg_packet_decrypt_worker+0x73c/0x780 drivers/net/wireguard/receive.c:506
process_one_work+0x3f6/0x960 kernel/workqueue.c:2307
worker_thread+0x616/0xa70 kernel/workqueue.c:2454
kthread+0x1bf/0x1e0 kernel/kthread.c:377
ret_from_fork+0x1f/0x30
value changed: 0xffff88815883e000 -> 0x0000000000000000
Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 26435 Comm: kworker/0:1 Not tainted 5.17.0-rc1-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: wg-crypt-wg2 wg_packet_decrypt_worker
Fixes: 4ea7e38696 ("dropmon: add ability to detect when hardware dropsrxpackets")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ->rtm_tos option is normally used to route packets based on both
the destination address and the DS field. However it's ignored for
IPv6 routes. Setting ->rtm_tos for IPv6 is thus invalid as the route
is going to work only on the destination address anyway, so it won't
behave as specified.
Suggested-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vladimir Oltean says:
====================
More aggressive DSA cleanup
This series deletes some code which is apparently not needed.
I've had these patches in my tree for a while, and testing on my boards
didn't reveal any issues.
Compared to the RFC v1 series, the only change is the addition of patch 3.
https://patchwork.kernel.org/project/netdevbpf/cover/20220107184842.550334-1-vladimir.oltean@nxp.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 2f1e8ea726 ("net: dsa: link interfaces with the DSA
master to get rid of lockdep warnings"), suggested by Cong Wang, the
DSA interfaces and their master have different dev->nested_level, which
makes netif_addr_lock() stop complaining about potentially recursive
locking on the same lock class.
So we no longer need DSA slave interfaces to have their own lockdep
class.
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 2f1e8ea726 ("net: dsa: link interfaces with the DSA
master to get rid of lockdep warnings"), suggested by Cong Wang, the
DSA interfaces and their master have different dev->nested_level, which
makes netif_addr_lock() stop complaining about potentially recursive
locking on the same lock class.
So we no longer need DSA masters to have their own lockdep class.
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are no legacy ports, DSA registers a devlink instance with ports
unconditionally for all switch drivers. Therefore, delete the old-style
ndo operations used for determining bridge forwarding domains.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
D. Wythe says:
====================
net/smc: Optimizing performance in short-lived scenarios
This patch set aims to optimizing performance of SMC in short-lived
links scenarios, which is quite unsatisfactory right now.
In our benchmark, we test it with follow scripts:
./wrk -c 10000 -t 4 -H 'Connection: Close' -d 20 http://smc-server
Current performance figures like that:
Running 20s test @ http://11.213.45.6
4 threads and 10000 connections
4956 requests in 20.06s, 3.24MB read
Socket errors: connect 0, read 0, write 672, timeout 0
Requests/sec: 247.07
Transfer/sec: 165.28KB
There are many reasons for this phenomenon, this patch set doesn't
solve it all though, but it can be well alleviated with it in.
Patch 1/5 (Make smc_tcp_listen_work() independent) :
Separate smc_tcp_listen_work() from smc_listen_work(), make them
independent of each other, the busy SMC handshake can not affect new TCP
connections visit any more. Avoid discarding a large number of TCP
connections after being overstock, which is undoubtedly raise the
connection establishment time.
Patch 2/5 (Limit SMC backlog connections):
Since patch 1 has separated smc_tcp_listen_work() from
smc_listen_work(), an unrestricted TCP accept have come into being. This
patch try to put a limit on SMC backlog connections refers to
implementation of TCP.
Patch 3/5 (Limit SMC visits when handshake workqueue congested):
Considering the complexity of SMC handshake right now, in short-lived
links scenarios, this may not be the main scenario of SMC though, it's
performance is still quite poor. This patch try to provide constraint on
SMC handshake when handshake workqueue congested, which is the sign of
SMC handshake stacking in our opinion.
Patch 4/5 (Dynamic control handshake limitation by socket options)
This patch allow applications dynamically control the ability of SMC
handshake limitation. Since SMC don't support set SMC socket option
before,
this patch also have to support SMC's owns socket options.
Patch 5/5 (Add global configure for handshake limitation by netlink)
This patch provides a way to get benefit of handshake limitation
without
modifying any code for applications, which is quite useful for most
existing applications.
After this patch set, performance figures like that:
Running 20s test @ http://11.213.45.6
4 threads and 10000 connections
693253 requests in 20.10s, 452.88MB read
Requests/sec: 34488.13
Transfer/sec: 22.53MB
That's a quite well performance improvement, about to 6 to 7 times in my
environment.
---
changelog:
v1 -> v2:
- fix compile warning
- fix invalid dependencies in kconfig
v2 -> v3:
- correct spelling mistakes
- fix useless variable declare
v3 -> v4
- make smc_tcp_ls_wq be static
v4 -> v5
- add dynamic control for SMC auto fallback by socket options
- add global configure for SMC auto fallback through netlink
v5 -> v6
- move auto fallback to net namespace scope
- remove auto fallback attribute in SMC_GEN_SYS_INFO
- add independent attributes for auto fallback
v6 -> v7
- fix wording and the naming issues, rename 'auto fallback' to handshake
limitation.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Although we can control SMC handshake limitation through socket options,
which means that applications who need it must modify their code. It's
quite troublesome for many existing applications. This patch modifies
the global default value of SMC handshake limitation through netlink,
providing a way to put constraint on handshake without modifies any code
for applications.
Suggested-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch aims to add dynamic control for SMC handshake limitation for
every smc sockets, in production environment, it is possible for the
same applications to handle different service types, and may have
different opinion on SMC handshake limitation.
This patch try socket options to complete it, since we don't have socket
option level for SMC yet, which requires us to implement it at the same
time.
This patch does the following:
- add new socket option level: SOL_SMC.
- add new SMC socket option: SMC_LIMIT_HS.
- provide getter/setter for SMC socket options.
Link: https://lore.kernel.org/all/20f504f961e1a803f85d64229ad84260434203bd.1644323503.git.alibuda@linux.alibaba.com/
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch intends to provide a mechanism to put constraint on SMC
connections visit according to the pressure of SMC handshake process.
At present, frequent visits will cause the incoming connections to be
backlogged in SMC handshake queue, raise the connections established
time. Which is quite unacceptable for those applications who base on
short lived connections.
There are two ways to implement this mechanism:
1. Put limitation after TCP established.
2. Put limitation before TCP established.
In the first way, we need to wait and receive CLC messages that the
client will potentially send, and then actively reply with a decline
message, in a sense, which is also a sort of SMC handshake, affect the
connections established time on its way.
In the second way, the only problem is that we need to inject SMC logic
into TCP when it is about to reply the incoming SYN, since we already do
that, it's seems not a problem anymore. And advantage is obvious, few
additional processes are required to complete the constraint.
This patch use the second way. After this patch, connections who beyond
constraint will not informed any SMC indication, and SMC will not be
involved in any of its subsequent processes.
Link: https://lore.kernel.org/all/1641301961-59331-1-git-send-email-alibuda@linux.alibaba.com/
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current implementation does not handling backlog semantics, one
potential risk is that server will be flooded by infinite amount
connections, even if client was SMC-incapable.
This patch works to put a limit on backlog connections, referring to the
TCP implementation, we divides SMC connections into two categories:
1. Half SMC connection, which includes all TCP established while SMC not
connections.
2. Full SMC connection, which includes all SMC established connections.
For half SMC connection, since all half SMC connections starts with TCP
established, we can achieve our goal by put a limit before TCP
established. Refer to the implementation of TCP, this limits will based
on not only the half SMC connections but also the full connections,
which is also a constraint on full SMC connections.
For full SMC connections, although we know exactly where it starts, it's
quite hard to put a limit before it. The easiest way is to block wait
before receive SMC confirm CLC message, while it's under protection by
smc_server_lgr_pending, a global lock, which leads this limit to the
entire host instead of a single listen socket. Another way is to drop
the full connections, but considering the cast of SMC connections, we
prefer to keep full SMC connections.
Even so, the limits of full SMC connections still exists, see commits
about half SMC connection below.
After this patch, the limits of backend connection shows like:
For SMC:
1. Client with SMC-capability can makes 2 * backlog full SMC connections
or 1 * backlog half SMC connections and 1 * backlog full SMC
connections at most.
2. Client without SMC-capability can only makes 1 * backlog half TCP
connections and 1 * backlog full TCP connections.
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In multithread and 10K connections benchmark, the backend TCP connection
established very slowly, and lots of TCP connections stay in SYN_SENT
state.
Client: smc_run wrk -c 10000 -t 4 http://server
the netstate of server host shows like:
145042 times the listen queue of a socket overflowed
145042 SYNs to LISTEN sockets dropped
One reason of this issue is that, since the smc_tcp_listen_work() shared
the same workqueue (smc_hs_wq) with smc_listen_work(), while the
smc_listen_work() do blocking wait for smc connection established. Once
the workqueue became congested, it's will block the accept() from TCP
listen.
This patch creates a independent workqueue(smc_tcp_ls_wq) for
smc_tcp_listen_work(), separate it from smc_listen_work(), which is
quite acceptable considering that smc_tcp_listen_work() runs very fast.
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
