Permalink
Commits on Oct 14, 2021
-
tcp: switch orphan_count to bare per-cpu counters
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 | --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 cf86a08 ("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. Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Stefan Bach <sfb@google.com> Cc: Neal Cardwell <ncardwell@google.com>
Commits on Oct 13, 2021
-
net: dsa: unregister cross-chip notifier after ds->ops->teardown
To be symmetric with the error unwind path of dsa_switch_setup(), call dsa_switch_unregister_notifier() after ds->ops->teardown. The implication is that ds->ops->teardown cannot emit cross-chip notifiers. For example, currently the dsa_tag_8021q_unregister() call from sja1105_teardown() does not propagate to the entire tree due to this reason. However I cannot find an actual issue caused by this, observed using code inspection. Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://lore.kernel.org/r/20211012123735.2545742-1-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-
marvell: octeontx2: build error: unknown type name 'u64'
Building an allmodconfig kernel arm64 kernel, the following build error shows up: In file included from drivers/crypto/marvell/octeontx2/cn10k_cpt.c:4: include/linux/soc/marvell/octeontx2/asm.h:38:15: error: unknown type name 'u64' 38 | static inline u64 otx2_atomic64_fetch_add(u64 incr, u64 *ptr) | ^~~ Include linux/types.h in asm.h so the compiler knows what the type 'u64' are. Fixes: af3826d ("octeontx2-pf: Use hardware register for CQE count") Signed-off-by: Anders Roxell <anders.roxell@linaro.org> Link: https://lore.kernel.org/r/20211013135743.3826594-1-anders.roxell@linaro.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> -
net: qed_debug: fix check of false (grc_param < 0) expression
The type of enum dbg_grc_params has the enumerator list starting from 0. When grc_param is declared by enum dbg_grc_params, (grc_param < 0) is always false. We should remove the check of this expression. Signed-off-by: Jean Sacren <sakiwit@gmail.com> Acked-by: Shai Malin <smalin@marvell.com> Link: https://lore.kernel.org/r/20211012074645.12864-1-sakiwit@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-
net: enetc: include ip6_checksum.h for csum_ipv6_magic
For those architectures which do not define_HAVE_ARCH_IPV6_CSUM, we need to include ip6_checksum.h which provides the csum_ipv6_magic() function. Fixes: fb8629e ("net: enetc: add support for software TSO") Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://lore.kernel.org/r/20211012121358.16641-1-ioana.ciornei@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-
ionic: no devlink_unregister if not registered
Don't try to unregister the devlink if it hasn't been registered yet. This bit of error cleanup code got missed in the recent devlink registration changes. Fixes: 7911c8b ("ionic: Move devlink registration to be last devlink command") Signed-off-by: Shannon Nelson <snelson@pensando.io> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Link: https://lore.kernel.org/r/20211012231520.72582-1-snelson@pensando.io Signed-off-by: Jakub Kicinski <kuba@kernel.org>
