Eric Dumazet 52085f9e7e tcp: switch orphan_count to bare per-cpu counters ... [ Upstream commit 19757cebf0 ] Use of percpu_counter structure to track count of orphaned sockets is causing problems on modern hosts with 256 cpus or more. Stefan Bach reported a serious spinlock contention in real workloads, that I was able to reproduce with a netfilter rule dropping incoming FIN packets. 53.56% server [kernel.kallsyms] [k] queued_spin_lock_slowpath | ---queued_spin_lock_slowpath | --53.51%--_raw_spin_lock_irqsave | --53.51%--__percpu_counter_sum tcp_check_oom | |--39.03%--__tcp_close | tcp_close | inet_release | inet6_release | sock_close | __fput | ____fput | task_work_run | exit_to_usermode_loop | do_syscall_64 | entry_SYSCALL_64_after_hwframe | __GI___libc_close | --14.48%--tcp_out_of_resources tcp_write_timeout tcp_retransmit_timer tcp_write_timer_handler tcp_write_timer call_timer_fn expire_timers __run_timers run_timer_softirq __softirqentry_text_start As explained in commit cf86a086a1 ("net/dst: use a smaller percpu_counter batch for dst entries accounting"), default batch size is too big for the default value of tcp_max_orphans (262144). But even if we reduce batch sizes, there would still be cases where the estimated count of orphans is beyond the limit, and where tcp_too_many_orphans() has to call the expensive percpu_counter_sum_positive(). One solution is to use plain per-cpu counters, and have a timer to periodically refresh this cache. Updating this cache every 100ms seems about right, tcp pressure state is not radically changing over shorter periods. percpu_counter was nice 15 years ago while hosts had less than 16 cpus, not anymore by current standards. v2: Fix the build issue for CONFIG_CRYPTO_DEV_CHELSIO_TLS=m, reported by kernel test robot <lkp@intel.com> Remove unused socket argument from tcp_too_many_orphans() Fixes: dd24c00191 ("net: Use a percpu_counter for orphan_count") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Stefan Bach <sfb@google.com> Cc: Neal Cardwell <ncardwell@google.com> Acked-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>		2023-04-19 16:57:23 +08:00
..
acpi	ACPI: tools: fix compilation error	2021-10-07 19:18:19 +02:00
asm-generic	asm-generic: build fixes for v5.15	2021-10-08 11:57:54 -07:00
clocksource
crypto	crypto: add patch for 5.15	2022-04-28 13:54:49 +08:00
drm
dt-bindings	pinctrl: starfive: jh7110: Correct the ioconfig register address and bit definitions	2023-03-16 14:25:09 +08:00
keys
kunit	kunit: fix kernel-doc warnings due to mismatched arg names	2021-10-06 17:54:07 -06:00
kvm	KVM: arm64: Fix PMU probe ordering	2021-09-20 12:43:34 +01:00
linux	kernel/sched: Fix sched_fork() access an invalid sched_task_group	2023-04-19 16:57:23 +08:00
math-emu
media	media: videobuf2: rework vb2_mem_ops API	2023-04-19 16:57:08 +08:00
memory	memory: renesas-rpc-if: Correct QSPI data transfer in Manual mode	2023-04-19 16:57:00 +08:00
misc
net	tcp: switch orphan_count to bare per-cpu counters	2023-04-19 16:57:23 +08:00
pcmcia
ras
rdma
scsi	scsi: core: Avoid leaving shost->last_reset with stale value if EH does not run	2023-04-19 16:56:53 +08:00
soc	Add a pm function for GPU	2022-07-01 15:01:36 +08:00
sound	[Audio: I2S] Support WM8960 one channel play and captrue	2022-08-11 02:37:26 -07:00
target
trace	block-5.15-2021-10-17	2021-10-17 19:25:20 -10:00
uapi	signal: Add SA_IMMUTABLE to ensure forced siganls do not get changed	2023-04-19 16:57:00 +08:00
vdso
video	v4l2: modify v4l2 compatible name	2022-08-31 09:16:30 +08:00
xen	xen/privcmd: drop "pages" parameter from xen_remap_pfn()	2021-10-05 08:20:27 +02:00