Re: Sets update

["Kerin Millar" <kfm@plushkava.net>]

On Tue, 23 Jul 2024, at 10:39 AM, Pablo Neira Ayuso wrote:
> On Tue, Jul 23, 2024 at 09:24:39AM +0200, Slavko wrote:
> [...]
>> (The host has >6 GB of free RAM, the list has ~1700 items)
>> 
>> Initially i blame IP duplicates in downloaded list of IPs, but that was
>> unrelated to this error. When i remove the first "add" and "delete"
>> lines from script (thus just one "add"), it works. After initial fill it
>> works with all three commands (add+delete+add) on already filled set
>> and even when i flush that set, it still works. Only first fill (after
>> boot) ends with that error.
>> 
>> When i delete and create that set, i got "Out of memory" again, the set
>> is defined as::
>> 
>>     table inet fw4 {
>> 	set myset {
>> 		type ipv4_addr
>> 		last counter
>> 		timeout 2d
>> 	}
>>     }
>> 
>> I tried to add "size" into it, but that doesn't help.
>> 
>> I roughly remember, that i read something about some memory limit in
>> container (i am not in container), but i am not able to find that
>> again to check if that is problem.
>> 
>> Please, what can cause that initial fill error, how i can debug/solve
>> it?
>
> Kernel is likely missing this patch:
>
> commit fa23e0d4b756d25829e124d6b670a4c6bbd4bf7e
> Author: Florian Westphal <fw@strlen.de>
> Date:   Wed May 8 14:52:47 2024 +0200
>
>     netfilter: nf_tables: allow clone callbacks to sleep 

Would it be safe to backport the following two commits to linux-6.6.y in isolation?

3c13725f43dcf43ad8a9bcd6a9f12add19a8f93e (bail out if stateful expression provides no .clone)
fa23e0d4b756d25829e124d6b670a4c6bbd4bf7e (allow clone callbacks to sleep)

-- 
Kerin Millar