Jacob Keller says:
====================
Intel Wired LAN Driver Updates for 2023-10-25 (ice)
This series extends the ice driver with basic support for the E830 device
line. It does not include support for all device features, but enables basic
functionality to load and pass traffic.
Alice adds the 200G speed and PHY types supported by E830 hardware.
Dan extends the DDP package logic to support the E830 package segment.
Paul adds the basic registers and macros used by E830 hardware, and adds
support for handling variable length link status information from firmware.
Pawel removes some redundant zeroing of the PCI IDs list, and extends the
list to include the E830 device IDs.
====================
Link: https://lore.kernel.org/r/20231025214157.1222758-1-jacob.e.keller@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kalle Valo says:
====================
wireless-next patches for v6.7
The third, and most likely the last, features pull request for v6.7.
Fixes all over and only few small new features.
Major changes:
iwlwifi
- more Multi-Link Operation (MLO) work
ath12k
- QCN9274: mesh support
ath11k
- firmware-2.bin container file format support
* tag 'wireless-next-2023-10-26' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next: (155 commits)
wifi: ray_cs: Remove unnecessary (void*) conversions
Revert "wifi: ath11k: call ath11k_mac_fils_discovery() without condition"
wifi: ath12k: Introduce and use ath12k_sta_to_arsta()
wifi: ath12k: fix htt mlo-offset event locking
wifi: ath12k: fix dfs-radar and temperature event locking
wifi: ath11k: fix gtk offload status event locking
wifi: ath11k: fix htt pktlog locking
wifi: ath11k: fix dfs radar event locking
wifi: ath11k: fix temperature event locking
wifi: ath12k: rename the sc naming convention to ab
wifi: ath12k: rename the wmi_sc naming convention to wmi_ab
wifi: ath11k: add firmware-2.bin support
wifi: ath11k: qmi: refactor ath11k_qmi_m3_load()
wifi: rtw89: cleanup firmware elements parsing
wifi: rt2x00: rework MT7620 PA/LNA RF calibration
wifi: rt2x00: rework MT7620 channel config function
wifi: rt2x00: improve MT7620 register initialization
MAINTAINERS: wifi: rt2x00: drop Helmut Schaa
wifi: wlcore: main: replace deprecated strncpy with strscpy
wifi: wlcore: boot: replace deprecated strncpy with strscpy
...
====================
Link: https://lore.kernel.org/r/20231026090411.B2426C433CB@smtp.kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf-next 2023-10-26
We've added 51 non-merge commits during the last 10 day(s) which contain
a total of 75 files changed, 5037 insertions(+), 200 deletions(-).
The main changes are:
1) Add open-coded task, css_task and css iterator support.
One of the use cases is customizable OOM victim selection via BPF,
from Chuyi Zhou.
2) Fix BPF verifier's iterator convergence logic to use exact states
comparison for convergence checks, from Eduard Zingerman,
Andrii Nakryiko and Alexei Starovoitov.
3) Add BPF programmable net device where bpf_mprog defines the logic
of its xmit routine. It can operate in L3 and L2 mode,
from Daniel Borkmann and Nikolay Aleksandrov.
4) Batch of fixes for BPF per-CPU kptr and re-enable unit_size checking
for global per-CPU allocator, from Hou Tao.
5) Fix libbpf which eagerly assumed that SHT_GNU_verdef ELF section
was going to be present whenever a binary has SHT_GNU_versym section,
from Andrii Nakryiko.
6) Fix BPF ringbuf correctness to fold smp_mb__before_atomic() into
atomic_set_release(), from Paul E. McKenney.
7) Add a warning if NAPI callback missed xdp_do_flush() under
CONFIG_DEBUG_NET which helps checking if drivers were missing
the former, from Sebastian Andrzej Siewior.
8) Fix missed RCU read-lock in bpf_task_under_cgroup() which was throwing
a warning under sleepable programs, from Yafang Shao.
9) Avoid unnecessary -EBUSY from htab_lock_bucket by disabling IRQ before
checking map_locked, from Song Liu.
10) Make BPF CI linked_list failure test more robust,
from Kumar Kartikeya Dwivedi.
11) Enable samples/bpf to be built as PIE in Fedora, from Viktor Malik.
12) Fix xsk starving when multiple xsk sockets were associated with
a single xsk_buff_pool, from Albert Huang.
13) Clarify the signed modulo implementation for the BPF ISA standardization
document that it uses truncated division, from Dave Thaler.
14) Improve BPF verifier's JEQ/JNE branch taken logic to also consider
signed bounds knowledge, from Andrii Nakryiko.
15) Add an option to XDP selftests to use multi-buffer AF_XDP
xdp_hw_metadata and mark used XDP programs as capable to use frags,
from Larysa Zaremba.
16) Fix bpftool's BTF dumper wrt printing a pointer value and another
one to fix struct_ops dump in an array, from Manu Bretelle.
* tag 'for-netdev' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (51 commits)
netkit: Remove explicit active/peer ptr initialization
selftests/bpf: Fix selftests broken by mitigations=off
samples/bpf: Allow building with custom bpftool
samples/bpf: Fix passing LDFLAGS to libbpf
samples/bpf: Allow building with custom CFLAGS/LDFLAGS
bpf: Add more WARN_ON_ONCE checks for mismatched alloc and free
selftests/bpf: Add selftests for netkit
selftests/bpf: Add netlink helper library
bpftool: Extend net dump with netkit progs
bpftool: Implement link show support for netkit
libbpf: Add link-based API for netkit
tools: Sync if_link uapi header
netkit, bpf: Add bpf programmable net device
bpf: Improve JEQ/JNE branch taken logic
bpf: Fold smp_mb__before_atomic() into atomic_set_release()
bpf: Fix unnecessary -EBUSY from htab_lock_bucket
xsk: Avoid starving the xsk further down the list
bpf: print full verifier states on infinite loop detection
selftests/bpf: test if state loops are detected in a tricky case
bpf: correct loop detection for iterators convergence
...
====================
Link: https://lore.kernel.org/r/20231026150509.2824-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The recent firmware interface change has added 2 counters in struct
rx_port_stats_ext. This caused 2 stray ethtool counters to be
displayed.
Since new counters are added from time to time, fix it so that the
ethtool logic will only display up to the maximum known counters.
These 2 counters are not used by production firmware yet.
Fixes: 754fbf604f ("bnxt_en: Update firmware interface to 1.10.2.171")
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20231026013231.53271-1-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Page pool code is compiled conditionally, but the operations
are part of the shared netlink family. We can handle this
by reporting empty list of pools or -EOPNOTSUPP / -ENOSYS
but the cleanest way seems to be removing the ops completely
at compilation time. That way user can see that the page
pool ops are not present using genetlink introspection.
Same way they'd check if the kernel is "new enough" to
support the ops.
Extend the specs with the ability to specify the config
condition under which op (and its policies, etc.) should
be hidden.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20231025162253.133159-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Paolo Abeni:
"Including fixes from WiFi and netfilter.
Most regressions addressed here come from quite old versions, with the
exceptions of the iavf one and the WiFi fixes. No known outstanding
reports or investigation.
Fixes to fixes:
- eth: iavf: in iavf_down, disable queues when removing the driver
Previous releases - regressions:
- sched: act_ct: additional checks for outdated flows
- tcp: do not leave an empty skb in write queue
- tcp: fix wrong RTO timeout when received SACK reneging
- wifi: cfg80211: pass correct pointer to rdev_inform_bss()
- eth: i40e: sync next_to_clean and next_to_process for programming
status desc
- eth: iavf: initialize waitqueues before starting watchdog_task
Previous releases - always broken:
- eth: r8169: fix data-races
- eth: igb: fix potential memory leak in igb_add_ethtool_nfc_entry
- eth: r8152: avoid writing garbage to the adapter's registers
- eth: gtp: fix fragmentation needed check with gso"
* tag 'net-6.6-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (43 commits)
iavf: in iavf_down, disable queues when removing the driver
vsock/virtio: initialize the_virtio_vsock before using VQs
net: ipv6: fix typo in comments
net: ipv4: fix typo in comments
net/sched: act_ct: additional checks for outdated flows
netfilter: flowtable: GC pushes back packets to classic path
i40e: Fix wrong check for I40E_TXR_FLAGS_WB_ON_ITR
gtp: fix fragmentation needed check with gso
gtp: uapi: fix GTPA_MAX
Fix NULL pointer dereference in cn_filter()
sfc: cleanup and reduce netlink error messages
net/handshake: fix file ref count in handshake_nl_accept_doit()
wifi: mac80211: don't drop all unprotected public action frames
wifi: cfg80211: fix assoc response warning on failed links
wifi: cfg80211: pass correct pointer to rdev_inform_bss()
isdn: mISDN: hfcsusb: Spelling fix in comment
tcp: fix wrong RTO timeout when received SACK reneging
r8152: Block future register access if register access fails
r8152: Rename RTL8152_UNPLUG to RTL8152_INACCESSIBLE
r8152: Check for unplug in r8153b_ups_en() / r8153c_ups_en()
...
When we configure the kernel command line with 'mitigations=off' and set
the sysctl knob 'kernel.unprivileged_bpf_disabled' to 0, the commit
bc5bc309db ("bpf: Inherit system settings for CPU security mitigations")
causes issues in the execution of `test_progs -t verifier`. This is
because 'mitigations=off' bypasses Spectre v1 and Spectre v4 protections.
Currently, when a program requests to run in unprivileged mode
(kernel.unprivileged_bpf_disabled = 0), the BPF verifier may prevent
it from running due to the following conditions not being enabled:
- bypass_spec_v1
- bypass_spec_v4
- allow_ptr_leaks
- allow_uninit_stack
While 'mitigations=off' enables the first two conditions, it does not
enable the latter two. As a result, some test cases in
'test_progs -t verifier' that were expected to fail to run may run
successfully, while others still fail but with different error messages.
This makes it challenging to address them comprehensively.
Moreover, in the future, we may introduce more fine-grained control over
CPU mitigations, such as enabling only bypass_spec_v1 or bypass_spec_v4.
Given the complexity of the situation, rather than fixing each broken test
case individually, it's preferable to skip them when 'mitigations=off' is
in effect and introduce specific test cases for the new 'mitigations=off'
scenario. For instance, we can introduce new BTF declaration tags like
'__failure__nospec', '__failure_nospecv1' and '__failure_nospecv4'.
In this patch, the approach is to simply skip the broken test cases when
'mitigations=off' is enabled. The result of `test_progs -t verifier` as
follows after this commit,
Before this commit
==================
- without 'mitigations=off'
- kernel.unprivileged_bpf_disabled = 2
Summary: 74/948 PASSED, 388 SKIPPED, 0 FAILED
- kernel.unprivileged_bpf_disabled = 0
Summary: 74/1336 PASSED, 0 SKIPPED, 0 FAILED <<<<
- with 'mitigations=off'
- kernel.unprivileged_bpf_disabled = 2
Summary: 74/948 PASSED, 388 SKIPPED, 0 FAILED
- kernel.unprivileged_bpf_disabled = 0
Summary: 63/1276 PASSED, 0 SKIPPED, 11 FAILED <<<< 11 FAILED
After this commit
=================
- without 'mitigations=off'
- kernel.unprivileged_bpf_disabled = 2
Summary: 74/948 PASSED, 388 SKIPPED, 0 FAILED
- kernel.unprivileged_bpf_disabled = 0
Summary: 74/1336 PASSED, 0 SKIPPED, 0 FAILED <<<<
- with this patch, with 'mitigations=off'
- kernel.unprivileged_bpf_disabled = 2
Summary: 74/948 PASSED, 388 SKIPPED, 0 FAILED
- kernel.unprivileged_bpf_disabled = 0
Summary: 74/948 PASSED, 388 SKIPPED, 0 FAILED <<<< SKIPPED
Fixes: bc5bc309db ("bpf: Inherit system settings for CPU security mitigations")
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Closes: https://lore.kernel.org/bpf/CAADnVQKUBJqg+hHtbLeeC2jhoJAWqnmRAzXW3hmUCNSV9kx4sQ@mail.gmail.com
Link: https://lore.kernel.org/bpf/20231025031144.5508-1-laoar.shao@gmail.com
Currently, it is not possible to specify custom flags when building
samples/bpf. The flags are defined in TPROGS_CFLAGS/TPROGS_LDFLAGS
variables, however, when trying to override those from the make command,
compilation fails.
For example, when trying to build with PIE:
$ make -C samples/bpf TPROGS_CFLAGS="-fpie" TPROGS_LDFLAGS="-pie"
This is because samples/bpf/Makefile updates these variables, especially
appends include paths to TPROGS_CFLAGS and these updates are overridden
by setting the variables from the make command.
This patch introduces variables TPROGS_USER_CFLAGS/TPROGS_USER_LDFLAGS
for this purpose, which can be set from the make command and their
values are propagated to TPROGS_CFLAGS/TPROGS_LDFLAGS.
Signed-off-by: Viktor Malik <vmalik@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/2d81100b830a71f0e72329cc7781edaefab75f62.1698213811.git.vmalik@redhat.com
There are two possible mismatched alloc and free cases in BPF memory
allocator:
1) allocate from cache X but free by cache Y with a different unit_size
2) allocate from per-cpu cache but free by kmalloc cache or vice versa
So add more WARN_ON_ONCE checks in free_bulk() and __free_by_rcu() to
spot these mismatched alloc and free early.
Signed-off-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20231021014959.3563841-1-houtao@huaweicloud.com
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains Netfilter updates for net-next. Mostly
nf_tables updates with two patches for connlabel and br_netfilter.
1) Rename function name to perform on-demand GC for rbtree elements,
and replace async GC in rbtree by sync GC. Patches from Florian Westphal.
2) Use commit_mutex for NFT_MSG_GETRULE_RESET to ensure that two
concurrent threads invoking this command do not underrun stateful
objects. Patches from Phil Sutter.
3) Use single hook to deal with IP and ARP packets in br_netfilter.
Patch from Florian Westphal.
4) Use atomic_t in netns->connlabel use counter instead of using a
spinlock, also patch from Florian.
5) Cleanups for stateful objects infrastructure in nf_tables.
Patches from Phil Sutter.
6) Flush path uses opaque set element offered by the iterator, instead of
calling pipapo_deactivate() which looks up for it again.
7) Set backend .flush interface always succeeds, make it return void
instead.
8) Add struct nft_elem_priv placeholder structure and use it by replacing
void * to pass opaque set element representation from backend to frontend
which defeats compiler type checks.
9) Shrink memory consumption of set element transactions, by reducing
struct nft_trans_elem object size and reducing stack memory usage.
10) Use struct nft_elem_priv also for set backend .insert operation too.
11) Carry reset flag in nft_set_dump_ctx structure, instead of passing it
as a function argument, from Phil Sutter.
netfilter pull request 23-10-25
* tag 'nf-next-23-10-25' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next:
netfilter: nf_tables: Carry reset boolean in nft_set_dump_ctx
netfilter: nf_tables: set->ops->insert returns opaque set element in case of EEXIST
netfilter: nf_tables: shrink memory consumption of set elements
netfilter: nf_tables: expose opaque set element as struct nft_elem_priv
netfilter: nf_tables: set backend .flush always succeeds
netfilter: nft_set_pipapo: no need to call pipapo_deactivate() from flush
netfilter: nf_tables: Carry reset boolean in nft_obj_dump_ctx
netfilter: nf_tables: nft_obj_filter fits into cb->ctx
netfilter: nf_tables: Carry s_idx in nft_obj_dump_ctx
netfilter: nf_tables: A better name for nft_obj_filter
netfilter: nf_tables: Unconditionally allocate nft_obj_filter
netfilter: nf_tables: Drop pointless memset in nf_tables_dump_obj
netfilter: conntrack: switch connlabels to atomic_t
br_netfilter: use single forward hook for ip and arp
netfilter: nf_tables: Add locking for NFT_MSG_GETRULE_RESET requests
netfilter: nf_tables: Introduce nf_tables_getrule_single()
netfilter: nf_tables: Open-code audit log call in nf_tables_getrule()
netfilter: nft_set_rbtree: prefer sync gc to async worker
netfilter: nft_set_rbtree: rename gc deactivate+erase function
====================
Link: https://lore.kernel.org/r/20231025212555.132775-1-pablo@netfilter.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
If the preferred lifetime was less than the minimum required lifetime,
ipv6_create_tempaddr would error out without creating any new address.
On my machine and network, this error happened immediately with the
preferred lifetime set to 1 second, after a few minutes with the
preferred lifetime set to 4 seconds, and not at all with the preferred
lifetime set to 5 seconds. During my investigation, I found a Stack
Exchange post from another person who seems to have had the same
problem: They stopped getting new addresses if they lowered the
preferred lifetime below 3 seconds, and they didn't really know why.
The preferred lifetime is a preference, not a hard requirement. The
kernel does not strictly forbid new connections on a deprecated address,
nor does it guarantee that the address will be disposed of the instant
its total valid lifetime expires. So rather than disable IPv6 privacy
extensions altogether if the minimum required lifetime swells above the
preferred lifetime, it is more in keeping with the user's intent to
increase the temporary address's lifetime to the minimum necessary for
the current network conditions.
With these fixes, setting the preferred lifetime to 3 or 4 seconds "just
works" because the extra fraction of a second is practically
unnoticeable. It's even possible to reduce the time before deprecation
to 1 or 2 seconds by also disabling duplicate address detection (setting
/proc/sys/net/ipv6/conf/*/dad_transmits to 0). I realize that that is a
pretty niche use case, but I know at least one person who would gladly
sacrifice performance and convenience to be sure that they are getting
the maximum possible level of privacy.
Link: https://serverfault.com/a/1031168/310447
Signed-off-by: Alex Henrie <alexhenrie24@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20231024212312.299370-3-alexhenrie24@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When the ipv6 stack output a GSO packet, if its gso_size is larger than
dst MTU, then all segments would be fragmented. However, it is possible
for a GSO packet to have a trailing segment with smaller actual size
than both gso_size as well as the MTU, which leads to an "atomic
fragment". Atomic fragments are considered harmful in RFC-8021. An
Existing report from APNIC also shows that atomic fragments are more
likely to be dropped even it is equivalent to a no-op [1].
Add an extra check in the GSO slow output path. For each segment from
the original over-sized packet, if it fits with the path MTU, then avoid
generating an atomic fragment.
Link: https://www.potaroo.net/presentations/2022-03-01-ipv6-frag.pdf [1]
Fixes: b210de4f8c ("net: ipv6: Validate GSO SKB before finish IPv6 processing")
Reported-by: David Wragg <dwragg@cloudflare.com>
Signed-off-by: Yan Zhai <yan@cloudflare.com>
Link: https://lore.kernel.org/r/90912e3503a242dca0bc36958b11ed03a2696e5e.1698156966.git.yan@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In iavf_down, we're skipping the scheduling of certain operations if
the driver is being removed. However, the IAVF_FLAG_AQ_DISABLE_QUEUES
request must not be skipped in this case, because iavf_close waits
for the transition to the __IAVF_DOWN state, which happens in
iavf_virtchnl_completion after the queues are released.
Without this fix, "rmmod iavf" takes half a second per interface that's
up and prints the "Device resources not yet released" warning.
Fixes: c8de44b577 ("iavf: do not process adminq tasks when __IAVF_IN_REMOVE_TASK is set")
Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Tested-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20231025183213.874283-1-jacob.e.keller@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
This patch contains two late Netfilter's flowtable fixes for net:
1) Flowtable GC pushes back packets to classic path in every GC run,
ie. every second. This is because NF_FLOW_HW_ESTABLISHED is only
used by sched/act_ct (never set) and IPS_SEEN_REPLY might be unset
by the time the flow is offloaded (this status bit is only reliable
in the sched/act_ct datapath).
2) sched/act_ct logic to push back packets to classic path to reevaluate
if UDP flow is unidirectional only applies if IPS_HW_OFFLOAD_BIT is
set on and no hardware offload request is pending to be handled.
From Vlad Buslov.
These two patches fixes two problems that were introduced in the
previous 6.5 development cycle.
* tag 'nf-23-10-25' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
net/sched: act_ct: additional checks for outdated flows
netfilter: flowtable: GC pushes back packets to classic path
====================
Link: https://lore.kernel.org/r/20231025100819.2664-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Once VQs are filled with empty buffers and we kick the host, it can send
connection requests. If the_virtio_vsock is not initialized before,
replies are silently dropped and do not reach the host.
virtio_transport_send_pkt() can queue packets once the_virtio_vsock is
set, but they won't be processed until vsock->tx_run is set to true. We
queue vsock->send_pkt_work when initialization finishes to send those
packets queued earlier.
Fixes: 0deab087b1 ("vsock/virtio: use RCU to avoid use-after-free on the_virtio_vsock")
Signed-off-by: Alexandru Matei <alexandru.matei@uipath.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/20231024191742.14259-1-alexandru.matei@uipath.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Mat Martineau says:
====================
mptcp: Features and fixes for v6.7
Patch 1 adds a configurable timeout for the MPTCP connection when all
subflows are closed, to support break-before-make use cases.
Patch 2 is a fix for a 1-byte error in rx data counters with MPTCP
fastopen connections.
Patch 3 is a minor code cleanup.
Patches 4 & 5 add handling of rcvlowat for MPTCP sockets, with a
prerequisite patch to use a common scaling ratio between TCP and MPTCP.
Patch 6 improves efficiency of memory copying in MPTCP transmit code.
Patch 7 refactors syncing of socket options from the MPTCP socket to
its subflows.
Patches 8 & 9 help the MPTCP packet scheduler perform well by changing
the handling of notsent_lowat in subflows and how available buffer space
is calculated for MPTCP-level sends.
====================
Link: https://lore.kernel.org/r/20231023-send-net-next-20231023-2-v1-0-9dc60939d371@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MPTCP protocol account for the data enqueued on all the subflows
to the main socket send buffer, while the send buffer auto-tuning
algorithm set the main socket send buffer size as the max size among
the subflows.
That causes bad performances when at least one subflow is sndbuf
limited, e.g. due to very high latency, as the MPTCP scheduler can't
even fill such buffer.
Change the send-buffer auto-tuning algorithm to compute the main socket
send buffer size as the sum of all the subflows buffer size.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231023-send-net-next-20231023-2-v1-9-9dc60939d371@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Any latency related tuning taking action at the subflow level does
not really affect the user-space, as only the main MPTCP socket is
relevant.
Anyway any limiting setting may foul the MPTCP scheduler, not being
able to fully use the subflow-level cwin, leading to very poor b/w
usage.
Enforce notsent_lowat to be a no-op on every subflow.
Note that TCP_NOTSENT_LOWAT is currently not supported, and properly
dealing with that will require more invasive changes.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231023-send-net-next-20231023-2-v1-8-9dc60939d371@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MPTCP protocol allow setting sk_rcvlowat, but the value there
is currently ignored.
Additionally, the default subflows sk_rcvlowat basically disables per
subflow delayed ack: the MPTCP protocol move the incoming data from the
subflows into the msk socket as soon as the TCP stacks invokes the subflow
data_ready callback. Later, when __tcp_ack_snd_check() takes action,
the subflow-level copied_seq matches rcv_nxt, and that mandate for an
immediate ack.
Let the mptcp receive path be aware of such threshold, explicitly tracking
the amount of data available to be ready and checking vs sk_rcvlowat in
mptcp_poll() and before waking-up readers.
Additionally implement the set_rcvlowat() callback, to properly handle
the rcvbuf auto-tuning on sk_rcvlowat changes.
Finally to properly handle delayed ack, force the subflow level threshold
to 0 and instead explicitly ask for an immediate ack when the msk level th
is not reached.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231023-send-net-next-20231023-2-v1-5-9dc60939d371@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MPTCP protocol allows sockets with no alive subflows to stay
in ESTABLISHED status for and user-defined timeout, to allow for
later subflows creation.
Currently such timeout is constant - TCP_TIMEWAIT_LEN. Let the
user-space configure them via a newly added sysctl, to better cope
with busy servers and simplify (make them faster) the relevant
pktdrill tests.
Note that the new know does not apply to orphaned MPTCP socket
waiting for the data_fin handshake completion: they always wait
TCP_TIMEWAIT_LEN.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231023-send-net-next-20231023-2-v1-1-9dc60939d371@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull ACPI fix from Rafael Wysocki:
"Unbreak the ACPI NFIT driver after a recent change that inadvertently
altered its behavior (Xiang Chen)"
* tag 'acpi-6.6-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI: NFIT: Install Notify() handler before getting NFIT table
