netdev - Re: [PATCH net-next 0/5] mlxsw: Add support for RED qevent "mark"

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

Message-ID: <20210120164508.6009dbbd@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Date:   Wed, 20 Jan 2021 16:45:08 -0800
From:   Jakub Kicinski <kuba@...nel.org>
To:     Ido Schimmel <idosch@...sch.org>
Cc:     netdev@...r.kernel.org, davem@...emloft.net, petrm@...dia.com,
        jiri@...dia.com, amcohen@...dia.com, mlxsw@...dia.com,
        Ido Schimmel <idosch@...dia.com>
Subject: Re: [PATCH net-next 0/5] mlxsw: Add support for RED qevent "mark"

On Wed, 20 Jan 2021 11:14:37 +0200 Ido Schimmel wrote:
> On Tue, Jan 19, 2021 at 02:22:55PM -0800, Jakub Kicinski wrote:
> > On Sun, 17 Jan 2021 10:02:18 +0200 Ido Schimmel wrote:  
> > > From: Ido Schimmel <idosch@...dia.com>
> > > 
> > > The RED qdisc currently supports two qevents: "early_drop" and "mark". The
> > > filters added to the block bound to the "early_drop" qevent are executed on
> > > packets for which the RED algorithm decides that they should be
> > > early-dropped. The "mark" filters are similarly executed on ECT packets
> > > that are marked as ECN-CE (Congestion Encountered).
> > > 
> > > A previous patchset has offloaded "early_drop" filters on Spectrum-2 and
> > > later, provided that the classifier used is "matchall", that the action
> > > used is either "trap" or "mirred", and a handful or further limitations.  
> > 
> > For early_drop trap or mirred makes obvious sense, no explanation
> > needed.
> > 
> > But for marked as a user I'd like to see a _copy_ of the packet, 
> > while the original continues on its marry way to the destination.
> > I'd venture to say that e.g. for a DCTCP deployment mark+trap is
> > unusable, at least for tracing, because it distorts the operation 
> > by effectively dropping instead of marking.
> > 
> > Am I reading this right?  
> 
> You get a copy of the packet as otherwise it will create a lot of
> problems (like you wrote).

Hm, so am I missing some background on semantics on TC_ACT_TRAP?
Or perhaps you use a different action code?

AFAICT the code in the kernel is:

struct sk_buff *tcf_qevent_handle(...

	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
	case TC_ACT_TRAP:
		__qdisc_drop(skb, to_free);
		*ret = __NET_XMIT_STOLEN;
		return NULL;

Having TRAP mean DROP makes sense for filters, but in case of qevents
shouldn't they be a no-op?

Looking at sch_red looks like TRAP being a no-op would actually give us
the expected behavior.

> > If that is the case and you really want to keep the mark+trap
> > functionality - I feel like at least better documentation is needed.
> > The current two liner should also be rewritten, quoting from patch 1:
> >   
> > > * - ``ecn_mark``
> > >   - ``drop``
> > >   - Traps ECN-capable packets that were marked with CE (Congestion
> > >     Encountered) code point by RED algorithm instead of being dropped  
> > 
> > That needs to say that the trap is for datagrams trapped by a qevent.
> > Otherwise "Traps ... instead of being dropped" is too much of a
> > thought-shortcut, marked packets are not dropped.
> > 
> > (I'd also think that trap is better documented next to early_drop,
> > let's look at it from the reader's perspective)  
> 
> How about:
> 
> "Traps a copy of ECN-capable packets that were marked with CE

I think "Traps copies" or "Traps the copy of .. packet"?
I'm not a native speaker but there seems to be a grammatical mix here.

> (Congestion Encountered) code point by RED algorithm instead of being
> dropped. The trap is enabled by attaching a filter with action 'trap' to

... instead of those copies being dropped.

> the 'mark' qevent of the RED qdisc."
>
> In addition, this output:
> 
> $ devlink trap show pci/0000:06:00.0 trap ecn_mark 
> pci/0000:06:00.0:
>   name ecn_mark type drop generic true action trap group buffer_drops
> 
> Can be converted to:
> 
> $ devlink trap show pci/0000:06:00.0 trap ecn_mark 
> pci/0000:06:00.0:
>   name ecn_mark type drop generic true action mirror group buffer_drops
> 
> "mirror: The packet is forwarded by the underlying device and a copy is sent to
> the CPU."
> 
> In this case the action is static and you cannot change it.

Oh yes, that's nice, I thought mirror in traps means mirror to another
port. Are there already traps which implement the mirroring / trapping
a clone? Quick grep yields nothing of substance.