As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
i40evf uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
ice uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
igb uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
ixgb uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
This also removes a problematic use of disable_irq() in
a context it is forbidden, as explained in commit
af3e0fcf78 ("8139too: Use disable_irq_nosync() in
rtl8139_poll_controller()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
lasts for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
fm10k uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
ixgbevf uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
ixgbe uses NAPI for TX completions, so we better let core
networking stack call the napi->poll() to avoid the capture.
Reported-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Song Liu <songliubraving@fb.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to allow NAPI drivers to no longer provide
ndo_poll_controller() method, as it has been proven problematic.
team driver must not look at its presence, but instead call
netpoll_poll_dev() which factorize the needed actions.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As diagnosed by Song Liu, ndo_poll_controller() can
be very dangerous on loaded hosts, since the cpu
calling ndo_poll_controller() might steal all NAPI
contexts (for all RX/TX queues of the NIC). This capture
can last for unlimited amount of time, since one
cpu is generally not able to drain all the queues under load.
It seems that all networking drivers that do use NAPI
for their TX completions, should not provide a ndo_poll_controller().
NAPI drivers have netpoll support already handled
in core networking stack, since netpoll_poll_dev()
uses poll_napi(dev) to iterate through registered
NAPI contexts for a device.
This patch allows netpoll_poll_dev() to process NAPI
contexts even for drivers not providing ndo_poll_controller(),
allowing for following patches in NAPI drivers.
Also we export netpoll_poll_dev() so that it can be called
by bonding/team drivers in following patches.
Reported-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Song Liu <songliubraving@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up until now, mlxsw tolerated firmware versions that weren't exactly
matching the required version, if the branch number matched. That
allowed the users to test various firmware versions as long as they were
on the right branch.
On the other hand, it made it impossible for mlxsw to put a hard lower
bound on a version that fixes all problems known to date. If a user had
a somewhat older FW version installed, mlxsw would start up just fine,
possibly performing non-optimally as it would use features that trigger
problematic behavior.
Therefore tweak the check to accept any FW version that is:
- on the same branch as the preferred version, and
- the same as or newer than the preferred version.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DECLARE_* not DEFINE_*
Fixes: 8360ed6745 ("RDS: IB: Use DEFINE_PER_CPU_SHARED_ALIGNED for rds_ib_stats")
Signed-off-by: David S. Miller <davem@davemloft.net>
- Fix OMAP Device Tree compatible strings to match DT
-----BEGIN PGP SIGNATURE-----
iQIzBAABCgAdFiEEdrbJNaO+IJqU8IdIUa+KL4f8d2EFAlunqzUACgkQUa+KL4f8
d2H0pA/7Bobyk3ENOmKag1KbF6N8cXVhh7RZ4UQ7zZW3ReFmOYcScA9UkFNgYbxx
LSW+0Tw7IjWbeTi43AxYMnOYIXCDcaqJma72XsOPXD8dS2YEv7wHthAxzVdenjUR
7CMA5QrsLZKrJk466iArip4k7VLeUDRW7eVPsE9UOL3jzcaRsfzXxUvA4+BWBl/X
ueSh2NGMD2e6yqAYIwstVaEeUNdjEcSz+vnk0m2UAvJM2ox4e5OBMF5FHucZ42KN
LFQbOBCqhCKJ26hudQ+Zc3purkvHVnKSZHqKA4IvMP5Rb+BjPH0sU6gpTidaxQ2x
S6uzYELA+IbTqlKlgH1W84aThqaMqwSOJzjoi0zAQcNvuaffblNyqLUQEoSsyDgj
2aVAr+A5Duxt2A8qckKWk5dy6TAgcE00ImGSIndferUw1WCj22xsTQc3J3xO46r0
IYT6TLsmDoNo913gFfB6t8OdvNWaUMLGABe89NnQFd2mfBo3HMbS14JJAOnH3V2K
cCSbOdj2jHhw7zBfPTts7qSQC3CJw2QluhOPLf9NBl0AHJZYuk30DsTCNX1GDkyq
m73cP0owXREOSWR3LpS8M3pbbu+Jd/JOWkPVRv3mgA/47DPqNbH1A4nXz8mdDqhQ
ao+44ri2D3hUeFsHsjc+oKYsgxuoPPKGssPIFwJU8dT925RZX/Q=
=5veh
-----END PGP SIGNATURE-----
Merge tag 'mfd-fixes-4.19' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd
Lee writes:
"MFD fixes for v4.19
- Fix Dialog DA9063 regulator constraints issue causing failure in
probe
- Fix OMAP Device Tree compatible strings to match DT"
* tag 'mfd-fixes-4.19' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd:
mfd: omap-usb-host: Fix dts probe of children
mfd: da9063: Fix DT probing with constraints
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQRTLbB6QfY48x44uB6AXGG7T9hjvgUCW6dGhgAKCRCAXGG7T9hj
vs1UAPwPSDmelfUus+P5ndRQZdK/iQWuRgQUe9Gd3RUVTfcQ7AEAljcN4/dSj7SB
hOgRlCp5WB1s5/vFF7z4jc2wtqvOPAk=
=8P9c
-----END PGP SIGNATURE-----
Merge tag 'for-linus-4.19d-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip
Juergen writes:
"xen:
Two small fixes for xen drivers."
* tag 'for-linus-4.19d-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen: issue warning message when out of grant maptrack entries
xen/x86/vpmu: Zero struct pt_regs before calling into sample handling code
-----BEGIN PGP SIGNATURE-----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=7uhx
-----END PGP SIGNATURE-----
Merge tag 'for-linus-20180922' of git://git.kernel.dk/linux-block
Jens writes:
"Just a single fix in this pull request, fixing a regression in
/proc/diskstats caused by the unification of timestamps."
* tag 'for-linus-20180922' of git://git.kernel.dk/linux-block:
block: use nanosecond resolution for iostat
Thomas writes:
"A set of fixes for x86:
- Resolve the kvmclock regression on AMD systems with memory
encryption enabled. The rework of the kvmclock memory allocation
during early boot results in encrypted storage, which is not
shareable with the hypervisor. Create a new section for this data
which is mapped unencrypted and take care that the later
allocations for shared kvmclock memory is unencrypted as well.
- Fix the build regression in the paravirt code introduced by the
recent spectre v2 updates.
- Ensure that the initial static page tables cover the fixmap space
correctly so early console always works. This worked so far by
chance, but recent modifications to the fixmap layout can -
depending on kernel configuration - move the relevant entries to a
different place which is not covered by the initial static page
tables.
- Address the regressions and issues which got introduced with the
recent extensions to the Intel Recource Director Technology code.
- Update maintainer entries to document reality"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mm: Expand static page table for fixmap space
MAINTAINERS: Add X86 MM entry
x86/intel_rdt: Add Reinette as co-maintainer for RDT
MAINTAINERS: Add Borislav to the x86 maintainers
x86/paravirt: Fix some warning messages
x86/intel_rdt: Fix incorrect loop end condition
x86/intel_rdt: Fix exclusive mode handling of MBA resource
x86/intel_rdt: Fix incorrect loop end condition
x86/intel_rdt: Do not allow pseudo-locking of MBA resource
x86/intel_rdt: Fix unchecked MSR access
x86/intel_rdt: Fix invalid mode warning when multiple resources are managed
x86/intel_rdt: Global closid helper to support future fixes
x86/intel_rdt: Fix size reporting of MBA resource
x86/intel_rdt: Fix data type in parsing callbacks
x86/kvm: Use __bss_decrypted attribute in shared variables
x86/mm: Add .bss..decrypted section to hold shared variables
Thomas writes:
"- Provide a strerror_r wrapper so lib/bpf can be built on systems
without _GNU_SOURCE
- Unbreak the man page generator when building out of tree"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf Documentation: Fix out-of-tree asciidoctor man page generation
tools lib bpf: Provide wrapper for strerror_r to build in !_GNU_SOURCE systems
Thomas writes:
"Make the EFI arm stub device tree loader default on to unbreak
existing EFI boot loaders which do not have DTB support."
* 'efi-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
efi/libstub/arm: default EFI_ARMSTUB_DTB_LOADER to y
Haiyang Zhang says:
====================
hv_netvsc: Support LRO/RSC in the vSwitch
The patch adds support for LRO/RSC in the vSwitch feature. It reduces
the per packet processing overhead by coalescing multiple TCP segments
when possible. The feature is enabled by default on VMs running on
Windows Server 2019 and later.
The patch set also adds ethtool command handler and documents.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Update document for LRO/RSC support, and the command line info to
change the setting.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the handler for LRO setting change, so that a user
can use ethtool command to enable / disable LRO feature.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
LRO/RSC in the vSwitch is a feature available in Windows Server 2019
hosts and later. It reduces the per packet processing overhead by
coalescing multiple TCP segments when possible. This patch adds netvsc
driver support for this feature.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So it should not fail with EPERM even though it is no longer implemented...
This is a fix for:
(userns)$ egrep ^Cap /proc/self/status
CapInh: 0000003fffffffff
CapPrm: 0000003fffffffff
CapEff: 0000003fffffffff
CapBnd: 0000003fffffffff
CapAmb: 0000003fffffffff
(userns)$ tcpdump -i usb_rndis0
tcpdump: WARNING: usb_rndis0: SIOCETHTOOL(ETHTOOL_GUFO) ioctl failed: Operation not permitted
Warning: Kernel filter failed: Bad file descriptor
tcpdump: can't remove kernel filter: Bad file descriptor
With this change it returns EOPNOTSUPP instead of EPERM.
See also https://github.com/the-tcpdump-group/libpcap/issues/689
Fixes: 08a00fea6d "net: Remove references to NETIF_F_UFO from ethtool."
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli says:
====================
net: dsa: b53: SGMII modes fixes
Here are two additional fixes that are required in order for SGMII to
work correctly. This was discovered with using a copper SFP which would
make us use SGMII mode, we would actually leave the HW configured in its
default mode: Fiber.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
In both 802.3z and SGMII modes we need to configure the MAC accordingly
to flip between Fiber and SGMII modes, and we need to read the MAC
status from the SGMII in-band control word.
Fixes: 0e01491de6 ("net: dsa: b53: Add SerDes support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Maths went wrong, to get 0x20, we need to do 0x1e + (x) * 2, not 0x18,
fix that offset so we access the correct registers. This would make us
not access the correct SerDes Digital control words, status would be
fine and so we would not be correctly flipping between Fiber and SGMII
modes resulting in incorrect status words being pulled into the SerDes
digital status register.
Fixes: 0e01491de6 ("net: dsa: b53: Add SerDes support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
PHYLINK takes care of filing the right information into
state->an_enabled, get rid of the read from the SerDes's BMCR register.
Fixes: 0e01491de6 ("net: dsa: b53: Add SerDes support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clang warns that the address of a pointer will always evaluated as true
in a boolean context.
net/decnet/dn_dev.c:1366:10: warning: address of array 'dev->name' will
always evaluate to 'true' [-Wpointer-bool-conversion]
dev->name ? dev->name : "???",
~~~~~^~~~ ~
1 warning generated.
Link: https://github.com/ClangBuiltLinux/linux/issues/116
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds ipv6 defragmentation tests to ip_defrag selftest,
to complement existing ipv4 tests.
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, ip[6]frag_high_thresh sysctl values in new namespaces are
hard-limited to those of the root/init ns.
There are at least two use cases when it would be desirable to
set the high_thresh values higher in a child namespace vs the global hard
limit:
- a security/ddos protection policy may lower the thresholds in the
root/init ns but allow for a special exception in a child namespace
- testing: a test running in a namespace may want to set these
thresholds higher in its namespace than what is in the root/init ns
The new behavior:
# ip netns add testns
# ip netns exec testns bash
# sysctl -w net.ipv4.ipfrag_high_thresh=9000000
net.ipv4.ipfrag_high_thresh = 9000000
# sysctl net.ipv4.ipfrag_high_thresh
net.ipv4.ipfrag_high_thresh = 9000000
# sysctl -w net.ipv6.ip6frag_high_thresh=9000000
net.ipv6.ip6frag_high_thresh = 9000000
# sysctl net.ipv6.ip6frag_high_thresh
net.ipv6.ip6frag_high_thresh = 9000000
The old behavior:
# ip netns add testns
# ip netns exec testns bash
# sysctl -w net.ipv4.ipfrag_high_thresh=9000000
net.ipv4.ipfrag_high_thresh = 9000000
# sysctl net.ipv4.ipfrag_high_thresh
net.ipv4.ipfrag_high_thresh = 4194304
# sysctl -w net.ipv6.ip6frag_high_thresh=9000000
net.ipv6.ip6frag_high_thresh = 9000000
# sysctl net.ipv6.ip6frag_high_thresh
net.ipv6.ip6frag_high_thresh = 4194304
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is similar to how ipv4 now behaves:
commit 0ff89efb52 ("ip: fail fast on IP defrag errors").
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clang warns when two declarations' section attributes don't match.
net/rds/ib_stats.c:40:1: warning: section does not match previous
declaration [-Wsection]
DEFINE_PER_CPU_SHARED_ALIGNED(struct rds_ib_statistics, rds_ib_stats);
^
./include/linux/percpu-defs.h:142:2: note: expanded from macro
'DEFINE_PER_CPU_SHARED_ALIGNED'
DEFINE_PER_CPU_SECTION(type, name,
PER_CPU_SHARED_ALIGNED_SECTION) \
^
./include/linux/percpu-defs.h:93:9: note: expanded from macro
'DEFINE_PER_CPU_SECTION'
extern __PCPU_ATTRS(sec) __typeof__(type) name;
\
^
./include/linux/percpu-defs.h:49:26: note: expanded from macro
'__PCPU_ATTRS'
__percpu __attribute__((section(PER_CPU_BASE_SECTION sec)))
\
^
net/rds/ib.h:446:1: note: previous attribute is here
DECLARE_PER_CPU(struct rds_ib_statistics, rds_ib_stats);
^
./include/linux/percpu-defs.h:111:2: note: expanded from macro
'DECLARE_PER_CPU'
DECLARE_PER_CPU_SECTION(type, name, "")
^
./include/linux/percpu-defs.h:87:9: note: expanded from macro
'DECLARE_PER_CPU_SECTION'
extern __PCPU_ATTRS(sec) __typeof__(type) name
^
./include/linux/percpu-defs.h:49:26: note: expanded from macro
'__PCPU_ATTRS'
__percpu __attribute__((section(PER_CPU_BASE_SECTION sec)))
\
^
1 warning generated.
The initial definition was added in commit ec16227e14 ("RDS/IB:
Infiniband transport") and the cache aligned definition was added in
commit e6babe4cc4 ("RDS/IB: Stats and sysctls") right after. The
definition probably should have been updated in net/rds/ib.h, which is
what this patch does.
Link: https://github.com/ClangBuiltLinux/linux/issues/114
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/fib_frontend.c: In function 'fib_info_nh_uses_dev':
net/ipv4/fib_frontend.c:322:6: error: unused variable 'ret' [-Werror=unused-variable]
cc1: all warnings being treated as errors
Fixes: 78f2756c5f ("net/ipv4: Move device validation to helper")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Ahern <dsahern@gmail.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp: switch to Early Departure Time model
In the early days, pacing has been implemented in sch_fq (FQ)
in a generic way :
- SO_MAX_PACING_RATE could be used by any sockets.
- TCP would vary effective pacing rate based on CWND*MSS/SRTT
- FQ would ensure delays between packets based on current
sk->sk_pacing_rate, but with some quantum based artifacts.
(inflating RPC tail latencies)
- BBR then tweaked the pacing rate in its various phases
(PROBE, DRAIN, ...)
This worked reasonably well, but had the side effect that TCP RTT
samples would be inflated by the sojourn time of the packets in FQ.
Also note that when FQ is not used and TCP wants pacing, the
internal pacing fallback has very different behavior, since TCP
emits packets at the time they should be sent (with unreasonable
assumptions about scheduling costs)
Van Jacobson gave a talk at Netdev 0x12 in Montreal, about letting
TCP (or applications for UDP messages) decide of the Earliest
Departure Time, instead of letting packet schedulers derive it
from pacing rate.
https://www.netdevconf.org/0x12/session.html?evolving-from-afap-teaching-nics-about-timehttps://www.files.netdevconf.org/d/46def75c2ef345809bbe/files/?p=/Evolving%20from%20AFAP%20%E2%80%93%20Teaching%20NICs%20about%20time.pdf
Recent additions in linux provided SO_TXTIME and a new ETF qdisc
supporting the new skb->tstamp role
This patch series converts TCP and FQ to the same model.
This might in the future allow us to relax tight TSQ limits
(if FQ is present in the output path), and thus lower
number of callbacks to tcp_write_xmit(), thanks to batching.
This will be followed by FQ change allowing SO_TXTIME support
so that QUIC servers can let the pacing being done in FQ (or
offloaded if network device permits)
For example, a TCP flow rated at 24Mbps now shows a more meaningful RTT
Before :
ESTAB 0 211408 10.246.7.151:41558 10.246.7.152:33723
cubic wscale:8,8 rto:203 rtt:2.195/0.084 mss:1448 rcvmss:536
advmss:1448 cwnd:20 ssthresh:20 bytes_acked:36897937
segs_out:25488 segs_in:12454 data_segs_out:25486
send 105.5Mbps lastsnd:1 lastrcv:12851 lastack:1
pacing_rate 24.0Mbps/24.0Mbps delivery_rate 22.9Mbps
busy:12851ms unacked:4 rcv_space:29200 notsent:205616 minrtt:0.026
After :
ESTAB 0 192584 10.246.7.151:61612 10.246.7.152:34375
cubic wscale:8,8 rto:201 rtt:0.165/0.129 mss:1448 rcvmss:536
advmss:1448 cwnd:20 ssthresh:20 bytes_acked:170755401
segs_out:117931 segs_in:57651 data_segs_out:117929
send 1404.1Mbps lastsnd:1 lastrcv:56915 lastack:1
pacing_rate 24.0Mbps/24.0Mbps delivery_rate 24.2Mbps
busy:56915ms unacked:4 rcv_space:29200 notsent:186792 minrtt:0.054
A nice side effect of this patch series is a reduction of max/p99
latencies of RPC workloads, since the FQ quantum no longer adds
artifact.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
With the earliest departure time model, we no longer plan
special casing TCP retransmits. We therefore remove dead
code (since most compilers understood skb_is_retransmit()
was false)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now TCP keeps track of tcp_wstamp_ns, recording the earliest
departure time of next packet, we can remove duplicate code
from tcp_internal_pacing()
This removes one ktime_get_tai_ns() call, and a divide.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP keeps track of tcp_wstamp_ns by itself, meaning sch_fq
no longer has to do it.
Thanks to this model, TCP can get more accurate RTT samples,
since pacing no longer inflates them.
This has the nice effect of removing some delays caused by FQ
quantum mechanism, causing inflated max/P99 latencies.
Also we might relax TCP Small Queue tight limits in the future,
since this new model allow TCP to build bigger batches, since
sch_fq (or a device with earliest departure time offload) ensure
these packets will be delivered on time.
Note that other protocols are not converted (they will probably
never be) so sch_fq has still support for SO_MAX_PACING_RATE
Tested:
Test showing FQ pacing quantum artifact for low-rate flows,
adding unexpected throttles for RPC flows, inflating max and P99 latencies.
The parameters chosen here are to show what happens typically when
a TCP flow has a reduced pacing rate (this can be caused by a reduced
cwin after few losses, or/and rtt above few ms)
MIBS="MIN_LATENCY,MEAN_LATENCY,MAX_LATENCY,P99_LATENCY,STDDEV_LATENCY"
Before :
$ netperf -H 10.246.7.133 -t TCP_RR -Cc -T6,6 -- -q 2000000 -r 100,100 -o $MIBS
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.246.7.133 () port 0 AF_INET : first burst 0 : cpu bind
Minimum Latency Microseconds,Mean Latency Microseconds,Maximum Latency Microseconds,99th Percentile Latency Microseconds,Stddev Latency Microseconds
19,82.78,5279,3825,482.02
After :
$ netperf -H 10.246.7.133 -t TCP_RR -Cc -T6,6 -- -q 2000000 -r 100,100 -o $MIBS
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.246.7.133 () port 0 AF_INET : first burst 0 : cpu bind
Minimum Latency Microseconds,Mean Latency Microseconds,Maximum Latency Microseconds,99th Percentile Latency Microseconds,Stddev Latency Microseconds
20,49.94,128,63,3.18
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Next patch will use tcp_wstamp_ns to feed internal
TCP pacing timer, so switch to CLOCK_TAI to share same base.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Switch internal TCP skb->skb_mstamp to skb->skb_mstamp_ns,
from usec units to nsec units.
Do not clear skb->tstamp before entering IP stacks in TX,
so that qdisc or devices can implement pacing based on the
earliest departure time instead of socket sk->sk_pacing_rate
Packets are fed with tcp_wstamp_ns, and following patch
will update tcp_wstamp_ns when both TCP and sch_fq switch to
the earliest departure time mechanism.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP will soon provide earliest departure time on TX skbs.
It needs to track this in a new variable.
tcp_mstamp_refresh() needs to update this variable, and
became too big to stay an inline.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP will soon provide per skb->tstamp with earliest departure time,
so that sch_fq does not have to determine departure time by looking
at socket sk_pacing_rate.
We chose in linux-4.19 CLOCK_TAI as the clock base for transports,
qdiscs, and NIC offloads.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are few places where TCP reads skb->skb_mstamp expecting
a value in usec unit.
skb->tstamp (aka skb->skb_mstamp) will soon store CLOCK_TAI nsec value.
Add tcp_skb_timestamp_us() to provide proper conversion when needed.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP pacing is either implemented in sch_fq or internally.
We have the goal of being able to offload pacing on the NICS.
TCP will soon provide per skb skb->tstamp as early departure time.
Like ETF in commit 25db26a913 ("net/sched: Introduce the ETF Qdisc")
we chose CLOCK_T as the clock base, so that TCP and pacers can share
a common clock, to get better RTT samples (without pacing artificially
inflating these samples).
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Salil Mehta says:
====================
Bug fixes, snall modifications & cleanup for HNS3 driver
This patch presents some bug fixes, small modifications and cleanups
to the HNS3 VF and PF driver.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes hclge_get_port_type which is redundant.
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Our VF has not implemented the ops for get_port_type. So when we executing
ethtool ethx cmd of VF, hns3_get_link_ksettings will return directly. And
we can not query anything.
To support get_link_ksettings for VF, this patch replaces get_port_type
with get_media_type. If the media type is HNAE3_MEDIA_TYPE_NONE,
hns3_get_link_ksettings will return link information of VF.
Fixes: 12f46bc1d4 ("net: hns3: Refine hns3_get_link_ksettings()")
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the ops of get_media_type support for VF.
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are already multiple types packets statistics for error packets,
it's unnecessary to print them, which may affect the rx performance if
print too many.
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For dma_mapping_error is unlikely happened, this patch adds unlikely for
dma_mapping_error check.
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When nic down, it firstly calls netif_tx_stop_all_queues(), then calls
napi_disable(). But napi_disable() will wait current napi_poll finish,
it may call netif_tx_wake_queue(). This patch fixes it by add nic state
checking.
Fixes: 424eb834a9 ("net: hns3: Unified HNS3 {VF|PF} Ethernet Driver for hip08 SoC")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet