[GLLUG] radvd leaking memory

[James Courtier-Dutton]

On Tue, 4 Feb 2020 at 17:29, Tim Woodall via GLLUG
<gllug at mailman.lug.org.uk> wrote:
>
> Hi,
>
> Last night radvd got killed by the oom killer
> Feb  4 02:32:11 firewall17 vmunix: [1762825.239631] [  pid  ]   uid  tgid total_vm      rss pgtables_bytes swapents oom_score_adj name
> Feb  4 02:32:11 firewall17 vmunix: [1762825.239671] [   3882]   105  3882    68490    19387   581632    48648             0 radvd
>
> When I look on another host I see that it's rather larger than I would
> have expected:
> 3247 radvd     20   0  177228  41016    556 S   0.0  24.0  16:07.26 radvd
>
> After a bounce it looks better:
>   1030 radvd     20   0    2444   1616   1476 S   0.0   0.9   0:00.01 radvd
>
>
> I cannot find anything about radvd having memory leaks. Does anyone know
> if this large memory footprint is expected (I'm running it on a VM with
> not much memory and I can increase that if necessary)?
>
> I can obviously restart it to avoid this issue. I wonder if it could be
> related to VPN interfaces that are constantly being created and removed.
>

Well its open source, so in theory you could diagnose the leak yourself.
clang has a very useful "-fsanitize=address"  option.
It allows the program to run pretty much at full speed, and then
reports memory leaks.