netdev - Re: [PATCH v4 net-next 00/19] inet: frags: bring rhashtables to IP defrag

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]

Date:   Wed, 30 May 2018 06:36:35 -0400
From:   Eric Dumazet <edumazet@...gle.com>
To:     Jesper Dangaard Brouer <brouer@...hat.com>
Cc:     Eric Dumazet <eric.dumazet@...il.com>, aring@...atatu.com,
        Tariq Toukan <tariqt@...lanox.com>,
        David Miller <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>, Florian Westphal <fw@...len.de>,
        Herbert Xu <herbert@...dor.apana.org.au>,
        Thomas Graf <tgraf@...g.ch>,
        Alexander Aring <alex.aring@...il.com>,
        Stefan Schmidt <stefan@....samsung.com>,
        Kirill Tkhai <ktkhai@...tuozzo.com>, moshe@...lanox.com,
        Eran Ben Elisha <eranbe@...lanox.com>,
        Rick Jones <rick.jones2@....com>
Subject: Re: [PATCH v4 net-next 00/19] inet: frags: bring rhashtables to IP defrag

On Wed, May 30, 2018 at 5:20 AM Jesper Dangaard Brouer <brouer@...hat.com>
wrote:

> On Mon, 28 May 2018 09:09:17 -0700
> Eric Dumazet <eric.dumazet@...il.com> wrote:

> > Tariq, here are my test results : No drops for me.
> >
> > # ./netperf -H 2607:f8b0:8099:e18:: -t UDP_STREAM
> > MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to
2607:f8b0:8099:e18:: () port 0 AF_INET6
> > Socket  Message  Elapsed      Messages
> > Size    Size     Time         Okay Errors   Throughput
> > bytes   bytes    secs            #      #   10^6bits/sec
> >
> > 212992   65507   10.00      202117      0    10592.00
> > 212992           10.00           0              0.00

> Hmm... Eric the above result show that ALL your UDP packets were dropped!
> You have 0 okay messages and 0.00 Mbit/s throughput.

> It needs to look like below (test on i40e NIC):

> $ netperf -t UDP_STREAM -H fee0:cafe::1
> MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to fee0:cafe::1 ()
port 0 AF_INET6 : histogram : demo
> Socket  Message  Elapsed      Messages
> Size    Size     Time         Okay Errors   Throughput
> bytes   bytes    secs            #      #   10^6bits/sec

> 212992   65507   10.00      186385      0    9767.08
> 212992           10.00      186385           9767.08


> If I manually instruct ip6tables to drop all UDP packets, then I get
> what you see... so, something on your test system are likely dropping
> your UDP packets, but letting regular netperf (TCP) control
> communication through.

> # ip6tables -I INPUT -p udp -j DROP

> $ netperf -t UDP_STREAM -H fee0:cafe::1
> MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to fee0:cafe::1 ()
port 0 AF_INET6 : histogram : demo
> Socket  Message  Elapsed      Messages
> Size    Size     Time         Okay Errors   Throughput
> bytes   bytes    secs            #      #   10^6bits/sec

> 212992   65507   10.00      182095      0    9542.41
> 212992           10.00           0              0.00



Right you are, for some reason I copied/pasted wrong results,
after _specifically_ filling up the frags to the memory limits,
when trying to reproduce 'bad numbers '

Here are the good ones, using latest David Miller net tree. ( plus
https://patchwork.ozlabs.org/patch/922528/  but that should not matter here)

llpaa23:/export/hda3/google/edumazet# ./netperf -H 2607:f8b0:8099:e18:: -t
UDP_STREAM
MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to
2607:f8b0:8099:e18:: () port 0 AF_INET6
Socket  Message  Elapsed      Messages
Size    Size     Time         Okay Errors   Throughput
bytes   bytes    secs            #      #   10^6bits/sec

212992   65507   10.00      216236      0    11331.89
212992           10.00      215068           11270.68


There are few drops because of the too small
/proc/sys/net/core/rmem_default  ( 212992 as seen in netperf output) for
these kind of stress.
( each 64KB datagram actually consumes half the budget ...)