Le 18/01/2017 à 02:01, Florian Westphal a écrit :
> This further refines the changes made to conntrack gc_worker in
> commit e0df8cae6c16 ("netfilter: conntrack: refine gc worker heuristics").
>
> The main idea of that change was to reduce the scan interval when evictions
> take place.
>
> However, on the reporters' setup, there are 1-2 million conntrack entries
> in total and roughly 8k new (and closing) connections per second.
>
> In this case we'll always evict at least one entry per gc cycle and scan
> interval is always at 1 jiffy because of this test:
>
> } else if (expired_count) {
> gc_work->next_gc_run /= 2U;
> next_run = msecs_to_jiffies(1);
>
> being true almost all the time.
>
> Given we scan ~10k entries per run its clearly wrong to reduce interval
> based on nonzero eviction count, it will only waste cpu cycles since a vast
> majorities of conntracks are not timed out.
>
> Thus only look at the ratio (scanned entries vs. evicted entries) to make
> a decision on whether to reduce or not.
>
> Because evictor is supposed to only kick in when system turns idle after
> a busy period, pick a high ratio -- this makes it 50%. We thus keep
> the idea of increasing scan rate when its likely that table contains many
> expired entries.
>
> In order to not let timed-out entries hang around for too long
> (important when using event logging, in which case we want to timely
> destroy events), we now scan the full table within at most
> GC_MAX_SCAN_JIFFIES (16 seconds) even in worst-case scenario where all
> timed-out entries sit in same slot.
>
> I tested this with a vm under synflood (with
> sysctl net.netfilter.nf_conntrack_tcp_timeout_syn_recv=3).
>
> While flood is ongoing, interval now stays at its max rate
> (GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV -> 125ms).
>
> With feedback from Nicolas Dichtel.
>
> Reported-by: Denys Fedoryshchenko <nuclearcat@xxxxxxxxxxxxxx>
> Cc: Nicolas Dichtel <nicolas.dichtel@xxxxxxxxx>
> Fixes: b87a2f9199ea82eaadc ("netfilter: conntrack: add gc worker to remove timed-out entries")
> Signed-off-by: Florian Westphal <fw@xxxxxxxxx>
Tested-by: Nicolas Dichtel <nicolas.dichtel@xxxxxxxxx>
Acked-by: Nicolas Dichtel <nicolas.dichtel@xxxxxxxxx>
Thank you,
Nicolas
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
