When no H2G transport is loaded, vsock currently routes all CIDs to the
G2H transport (commit 65b422d9b6 ("vsock: forward all packets to the
host when no H2G is registered"). Extend that existing behavior: when
an H2G transport is loaded but does not claim a given CID, the
connection falls back to G2H in the same way.
This matters in environments like Nitro Enclaves, where an instance may
run nested VMs via vhost-vsock (H2G) while also needing to reach sibling
enclaves at higher CIDs through virtio-vsock-pci (G2H). With the old
code, any CID > 2 was unconditionally routed to H2G when vhost was
loaded, making those enclaves unreachable without setting
VMADDR_FLAG_TO_HOST explicitly on every connect.
Requiring every application to set VMADDR_FLAG_TO_HOST creates friction:
tools like socat, iperf, and others would all need to learn about it.
The flag was introduced 6 years ago and I am still not aware of any tool
that supports it. Even if there was support, it would be cumbersome to
use. The most natural experience is a single CID address space where H2G
only wins for CIDs it actually owns, and everything else falls through to
G2H, extending the behavior that already exists when H2G is absent.
To give user space at least a hint that the kernel applied this logic,
automatically set the VMADDR_FLAG_TO_HOST on the remote address so it
can determine the path taken via getpeername().
Add a per-network namespace sysctl net.vsock.g2h_fallback (default 1).
At 0 it forces strict routing: H2G always wins for CID > VMADDR_CID_HOST,
or ENODEV if H2G is not loaded.
Signed-off-by: Alexander Graf <graf@amazon.com>
Tested-by: syzbot@syzkaller.appspotmail.com
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://patch.msgid.link/20260304230027.59857-1-graf@amazon.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add documentation for the vsock per-namespace sysctls (`ns_mode` and
`child_ns_mode`) to Documentation/admin-guide/sysctl/net.rst.
These sysctls were introduced by commit eafb64f40c ("vsock: add
netns to vsock core").
Document the two namespace modes (`global` and `local`), the
inheritance behavior of `child_ns_mode`, and the restriction preventing
local namespaces from setting `child_ns_mode` to `global`.
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://patch.msgid.link/20260216163147.236844-1-sgarzare@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
NETDEV_RSS_KEY_LEN has been set to 52 bytes in 2014, until now.
Jakub suggested we bump the size to 128 bytes or more.
Some drivers (like idpf) were already working around the core limit.
Since this change might cause some issues in admin scripts,
bump it directly to 256 in one go.
tjbp26:~# cat /proc/sys/net/core/netdev_rss_key | wc -c
768
tjbp26:~# ethtool -x eth1
RX flow hash indirection table for eth1 with 32 RX ring(s):
...
RSS hash key:
fe:16:5b:2f:93:85:c2:c9:c1:ef:bd:60:c6:e0:2b:99:4d:bf:b7:14:c8:1e:8d:cb:31:17:51:da:55:eb:91:d9:9e:f9:89:9b:44:a1:dc:08:72:3a:b3:d6:31:86:9a:fe:02:3a:0d:eb:a1:7c:f5:a3:51:3b:08:56:c9:3f:71:69:01:ba:70:38
RSS hash function:
toeplitz: on
xor: off
crc32: off
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/netdev/20260122075206.504ec591@kernel.org/
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20260122190349.2771064-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In blamed commit, I added a check against the temporary queue
built in __dev_xmit_skb(). Idea was to drop packets early,
before any spinlock was acquired.
if (unlikely(defer_count > READ_ONCE(q->limit))) {
kfree_skb_reason(skb, SKB_DROP_REASON_QDISC_DROP);
return NET_XMIT_DROP;
}
It turned out that HTB Qdisc has a zero q->limit.
HTB limits packets on a per-class basis.
Some of our tests became flaky.
Add a new sysctl : net.core.qdisc_max_burst to control
how many packets can be stored in the temporary lockless queue.
Also add a new QDISC_BURST_DROP drop reason to better diagnose
future issues.
Thanks Neal !
Fixes: 100dfa74ca ("net: dev_queue_xmit() llist adoption")
Reported-and-bisected-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Link: https://patch.msgid.link/20260107104159.3669285-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
If a socket has sk->sk_bypass_prot_mem flagged, the socket opts out
of the global protocol memory accounting.
Let's control the flag by a new sysctl knob.
The flag is written once during socket(2) and is inherited to child
sockets.
Tested with a script that creates local socket pairs and send()s a
bunch of data without recv()ing.
Setup:
# mkdir /sys/fs/cgroup/test
# echo $$ >> /sys/fs/cgroup/test/cgroup.procs
# sysctl -q net.ipv4.tcp_mem="1000 1000 1000"
# ulimit -n 524288
Without net.core.bypass_prot_mem, charged to tcp_mem & memcg
# python3 pressure.py &
# cat /sys/fs/cgroup/test/memory.stat | grep sock
sock 22642688 <-------------------------------------- charged to memcg
# cat /proc/net/sockstat| grep TCP
TCP: inuse 2006 orphan 0 tw 0 alloc 2008 mem 5376 <-- charged to tcp_mem
# ss -tn | head -n 5
State Recv-Q Send-Q Local Address:Port Peer Address:Port
ESTAB 2000 0 127.0.0.1:34479 127.0.0.1:53188
ESTAB 2000 0 127.0.0.1:34479 127.0.0.1:49972
ESTAB 2000 0 127.0.0.1:34479 127.0.0.1:53868
ESTAB 2000 0 127.0.0.1:34479 127.0.0.1:53554
# nstat | grep Pressure || echo no pressure
TcpExtTCPMemoryPressures 1 0.0
With net.core.bypass_prot_mem=1, charged to memcg only:
# sysctl -q net.core.bypass_prot_mem=1
# python3 pressure.py &
# cat /sys/fs/cgroup/test/memory.stat | grep sock
sock 2757468160 <------------------------------------ charged to memcg
# cat /proc/net/sockstat | grep TCP
TCP: inuse 2006 orphan 0 tw 0 alloc 2008 mem 0 <- NOT charged to tcp_mem
# ss -tn | head -n 5
State Recv-Q Send-Q Local Address:Port Peer Address:Port
ESTAB 111000 0 127.0.0.1:36019 127.0.0.1:49026
ESTAB 110000 0 127.0.0.1:36019 127.0.0.1:45630
ESTAB 110000 0 127.0.0.1:36019 127.0.0.1:44870
ESTAB 111000 0 127.0.0.1:36019 127.0.0.1:45274
# nstat | grep Pressure || echo no pressure
no pressure
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Roman Gushchin <roman.gushchin@linux.dev>
Link: https://patch.msgid.link/20251014235604.3057003-4-kuniyu@google.com
This will add eBPF JIT support to the 32-bit ARCv2 processors. The
implementation is qualified by running the BPF tests on a Synopsys HSDK
board with "ARC HS38 v2.1c at 500 MHz" as the 4-core CPU.
The test_bpf.ko reports 2-10 fold improvements in execution time of its
tests. For instance:
test_bpf: #33 tcpdump port 22 jited:0 704 1766 2104 PASS
test_bpf: #33 tcpdump port 22 jited:1 120 224 260 PASS
test_bpf: #141 ALU_DIV_X: 4294967295 / 4294967295 = 1 jited:0 238 PASS
test_bpf: #141 ALU_DIV_X: 4294967295 / 4294967295 = 1 jited:1 23 PASS
test_bpf: #776 JMP32_JGE_K: all ... magnitudes jited:0 2034681 PASS
test_bpf: #776 JMP32_JGE_K: all ... magnitudes jited:1 1020022 PASS
Deployment and structure
------------------------
The related codes are added to "arch/arc/net":
- bpf_jit.h -- The interface that a back-end translator must provide
- bpf_jit_core.c -- Knows how to handle the input eBPF byte stream
- bpf_jit_arcv2.c -- The back-end code that knows the translation logic
The bpf_int_jit_compile() at the end of bpf_jit_core.c is the entrance
to the whole process. Normally, the translation is done in one pass,
namely the "normal pass". In case some relocations are not known during
this pass, some data (arc_jit_data) is allocated for the next pass to
come. This possible next (and last) pass is called the "extra pass".
1. Normal pass # The necessary pass
1a. Dry run # Get the whole JIT length, epilogue offset, etc.
1b. Emit phase # Allocate memory and start emitting instructions
2. Extra pass # Only needed if there are relocations to be fixed
2a. Patch relocations
Support status
--------------
The JIT compiler supports BPF instructions up to "cpu=v4". However, it
does not yet provide support for:
- Tail calls
- Atomic operations
- 64-bit division/remainder
- BPF_PROBE_MEM* (exception table)
The result of "test_bpf" test suite on an HSDK board is:
hsdk-lnx# insmod test_bpf.ko test_suite=test_bpf
test_bpf: Summary: 863 PASSED, 186 FAILED, [851/851 JIT'ed]
All the failing test cases are due to the ones that were not JIT'ed.
Categorically, they can be represented as:
.-----------.------------.-------------.
| test type | opcodes | # of cases |
|-----------+------------+-------------|
| atomic | 0xC3, 0xDB | 149 |
| div64 | 0x37, 0x3F | 22 |
| mod64 | 0x97, 0x9F | 15 |
`-----------^------------+-------------|
| (total) 186 |
`-------------'
Setup: build config
-------------------
The following configs must be set to have a working JIT test:
CONFIG_BPF_JIT=y
CONFIG_BPF_JIT_ALWAYS_ON=y
CONFIG_TEST_BPF=m
The following options are not necessary for the tests module,
but are good to have:
CONFIG_DEBUG_INFO=y # prerequisite for below
CONFIG_DEBUG_INFO_BTF=y # so bpftool can generate vmlinux.h
CONFIG_FTRACE=y #
CONFIG_BPF_SYSCALL=y # all these options lead to
CONFIG_KPROBE_EVENTS=y # having CONFIG_BPF_EVENTS=y
CONFIG_PERF_EVENTS=y #
Some BPF programs provide data through /sys/kernel/debug:
CONFIG_DEBUG_FS=y
arc# mount -t debugfs debugfs /sys/kernel/debug
Setup: elfutils
---------------
The libdw.{so,a} library that is used by pahole for processing
the final binary must come from elfutils 0.189 or newer. The
support for ARCv2 [1] has been added since that version.
[1]
https://sourceware.org/git/?p=elfutils.git;a=commit;h=de3d46b3e7
Setup: pahole
-------------
The line below in linux/scripts/Makefile.btf must be commented out:
pahole-flags-$(call test-ge, $(pahole-ver), 121) += --btf_gen_floats
Or else, the build will fail:
$ make V=1
...
BTF .btf.vmlinux.bin.o
pahole -J --btf_gen_floats \
-j --lang_exclude=rust \
--skip_encoding_btf_inconsistent_proto \
--btf_gen_optimized .tmp_vmlinux.btf
Complex, interval and imaginary float types are not supported
Encountered error while encoding BTF.
...
BTFIDS vmlinux
./tools/bpf/resolve_btfids/resolve_btfids vmlinux
libbpf: failed to find '.BTF' ELF section in vmlinux
FAILED: load BTF from vmlinux: No data available
This is due to the fact that the ARC toolchains generate
"complex float" DIE entries in libgcc and at the moment, pahole
can't handle such entries.
Running the tests
-----------------
host$ scp /bld/linux/lib/test_bpf.ko arc:
arc # sysctl net.core.bpf_jit_enable=1
arc # insmod test_bpf.ko test_suite=test_bpf
...
test_bpf: #1048 Staggered jumps: JMP32_JSLE_X jited:1 697811 PASS
test_bpf: Summary: 863 PASSED, 186 FAILED, [851/851 JIT'ed]
Acknowledgments
---------------
- Claudiu Zissulescu for his unwavering support
- Yuriy Kolerov for testing and troubleshooting
- Vladimir Isaev for the pahole workaround
- Sergey Matyukevich for paving the road by adding the interpreter support
Signed-off-by: Shahab Vahedi <shahab@synopsys.com>
Link: https://lore.kernel.org/r/20240430145604.38592-1-list+bpf@vahedi.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch adds /proc/sys/net/core/mem_pcpu_rsv sysctl file,
to make SK_MEMORY_PCPU_RESERV tunable.
Commit 3cd3399dd7 ("net: implement per-cpu reserves for
memory_allocated") introduced per-cpu forward alloc cache:
"Implement a per-cpu cache of +1/-1 MB, to reduce number
of changes to sk->sk_prot->memory_allocated, which
would otherwise be cause of false sharing."
sk_prot->memory_allocated points to global atomic variable:
atomic_long_t tcp_memory_allocated ____cacheline_aligned_in_smp;
If increasing the per-cpu cache size from 1MB to e.g. 16MB,
changes to sk->sk_prot->memory_allocated can be further reduced.
Performance may be improved on system with many cores.
Signed-off-by: Adam Li <adamli@os.amperecomputing.com>
Reviewed-by: Christoph Lameter (Ampere) <cl@linux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For many years, /proc/sys/net/core/optmem_max default value
on a 64bit kernel has been 20 KB.
Regular usage of TCP tx zerocopy needs a bit more.
Google has used 128KB as the default value for 7 years without
any problem.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The net.core.txrehash documentation mentions this knob is for listening
sockets only, while sk_rethink_txhash can be called on SYN and RTO
retransmits on all TCP sockets.
Remove the listening socket part.
Signed-off-by: Antoine Tenart <atenart@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
If RPS is enabled, this allows configuring a default rps
mask, which is effective since receive queue creation time.
A default RPS mask allows the system admin to ensure proper
isolation, avoiding races at network namespace or device
creation time.
The default RPS mask is initially empty, and can be
modified via a newly added sysctl entry.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stephen Rothwell reported htmldocs warning when merging net-next tree:
Documentation/admin-guide/sysctl/net.rst:37: WARNING: Malformed table.
Text in column margin in table line 4.
========= =================== = ========== ==================
Directory Content Directory Content
========= =================== = ========== ==================
802 E802 protocol mptcp Multipath TCP
appletalk Appletalk protocol netfilter Network Filter
ax25 AX25 netrom NET/ROM
bridge Bridging rose X.25 PLP layer
core General parameter tipc TIPC
ethernet Ethernet protocol unix Unix domain sockets
ipv4 IP version 4 x25 X.25 protocol
ipv6 IP version 6
========= =================== = ========== ==================
The warning above is caused by cells in second "Content" column of
/proc/sys/net subdirectory table which are in column margin.
Align these cells against the column header to fix the warning.
Link: https://lore.kernel.org/linux-next/20220823134905.57ed08d5@canb.auug.org.au/
Fixes: 1202cdd665 ("Remove DECnet support from kernel")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20220824035804.204322-1-bagasdotme@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While reading netdev_max_backlog, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.
While at it, we remove the unnecessary spaces in the doc.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
DECnet is an obsolete network protocol that receives more attention
from kernel janitors than users. It belongs in computer protocol
history museum not in Linux kernel.
It has been "Orphaned" in kernel since 2010. The iproute2 support
for DECnet was dropped in 5.0 release. The documentation link on
Sourceforge says it is abandoned there as well.
Leave the UAPI alone to keep userspace programs compiling.
This means that there is still an empty neighbour table
for AF_DECNET.
The table of /proc/sys/net entries was updated to match
current directories and reformatted to be alphabetical.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: David Ahern <dsahern@kernel.org>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
A description is missing for the net.core.high_order_alloc_disable
option in admin-guide/sysctl/net.rst ; add it. The above sysctl option
was introduced by commit ce27ec6064 ("net: add high_order_alloc_disable
sysctl/static key").
Thanks to Eric for running again the benchmark cited in the above
commit, showing this knob is now mostly of historical importance.
Signed-off-by: Antoine Tenart <atenart@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220707080245.180525-1-atenart@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
commit 68822bdf76 ("net: generalize skb freeing
deferral to per-cpu lists") added another per-cpu
cache of skbs. It was expected to be small,
and an IPI was forced whenever the list reached 128
skbs.
We might need to be able to control more precisely
queue capacity and added latency.
An IPI is generated whenever queue reaches half capacity.
Default value of the new limit is 64.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update Documentation/admin-guide/sysctl/net.rst with txrehash usage
description.
Signed-off-by: Akhmat Karakotov <hmukos@yandex-team.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
