netdev - Re: [PATCH net-next 0/2] sctp: fully support memory accounting

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CADvbK_fcGR_U-URF-5N1i=nK=u7Xxa+tkfWeb5hTzMPFukHysg@mail.gmail.com>
Date:   Fri, 5 Apr 2019 12:49:37 +0800
From:   Xin Long <lucien.xin@...il.com>
To:     David Miller <davem@...emloft.net>
Cc:     network dev <netdev@...r.kernel.org>, linux-sctp@...r.kernel.org,
        Marcelo Ricardo Leitner <marcelo.leitner@...il.com>,
        Neil Horman <nhorman@...driver.com>,
        Matteo Croce <mcroce@...hat.com>,
        Vladis Dronov <vdronov@...hat.com>
Subject: Re: [PATCH net-next 0/2] sctp: fully support memory accounting

On Fri, Apr 5, 2019 at 1:38 AM David Miller <davem@...emloft.net> wrote:
>
> From: Xin Long <lucien.xin@...il.com>
> Date: Thu, 4 Apr 2019 17:45:07 +0800
>
> > We're thinking about adding some sctp tests into selftests,
> > but these ones are not good to be the first one, I think.
> > sysctl_sockets_memctl.sh is a multi-host test, even netns can't work for it.
>
> You can't set per-netns memory controls appropriately to make the
> tests work?
static atomic_long_t sctp_memory_allocated;

struct proto sctp_prot = {
.memory_allocated = &sctp_memory_allocated,

sysctl limitation is per-netns, but the counter 'sctp_memory_allocated' is not.
It means other netns buffer allocation will affect the current netns.

>
> > Each subcomponent seems to have its own test case in other git repo,
> > can I ask what kind of tests should I put into kernel selftests in
> > the future?
>
> It would be ideal that we have a decent base set of SCTP tests and
> then when bug fixes are added, we get a unit test with the fix.
>
> I guess a good baseline would be testing basic comminucation between
> SCTP sockets in different netns, and then adding tests for setting
> the various socket options and making sure the socket option had the
> desired effect.
got your point. thanks