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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CA+icZUW3vDp3HzjoXWtVz_zRa0TwACc1p2EV0c4W=jeWScZUuQ@mail.gmail.com>
Date:	Thu, 31 Dec 2015 15:06:18 +0100
From:	Sedat Dilek <sedat.dilek@...il.com>
To:	Guillaume Nault <g.nault@...halink.fr>
Cc:	David Miller <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [net-next] ppp: rtnetlink device handling

On Thu, Dec 31, 2015 at 1:58 PM, Guillaume Nault <g.nault@...halink.fr> wrote:
> On Thu, Dec 31, 2015 at 12:01:40PM +0100, Sedat Dilek wrote:
>> On Thu, Dec 31, 2015 at 11:41 AM, Guillaume Nault <g.nault@...halink.fr> wrote:
>> > On Thu, Dec 31, 2015 at 08:46:59AM +0100, Sedat Dilek wrote:
>> >> Hi Guillaume,
>> >>
>> >> can you explain why you moved ppp to rtnetlink device handling?
>> >> Benefits, etc.?
>> >>
>> > The objective is to bring all the flexibility of rtnetlink device
>> > creation to ppp interfaces. This is particularly useful for Network
>> > Access Servers which need to define device properties at creation
>> > time. My use case includes setting device name and netns, which is
>> > impossible with the ioctl-based interface without generating transient
>> > effects.
>> >
>>
>> What do you mean by "my use-case"?
>> Example?
>>
> A NAS that terminates connections of PPP clients (usually transported
> over PPPoE or L2TP). Clients may have different routing needs so each
> PPP device is created in the appropriate network namespace. Without
> rtnl the PPP device has to be created in the PPPoE/L2TP server's
> namespace and then moved to the client's one. With rtnl, the PPP device
> can be directly created in the right namespace.

Just off-topic...

Can you tell something to optimize my PPP Internet connection?

#1: BUFFERBLOAT

>From bufferbloat project I have enabled htb and fq_codel network-scheduler.

#2: TXQUEUELEN

What about increasing txqueuelen value?
Default for PPP is "3".

What about the calculation in "RE: ppp0 and it's txqueuelen with HSDPA
gsm modems"?

[ Q ]
> What would be "optimal" value for ppp0 device's txqueuelen,
> when it is used with HSDPA gsm modem for link which downlink
> speed is 1Mbit/s and uplink speed is 384kbit/s and ping
> ~80..100 ms? Is the default value 3 just OK?

[ A ]
"I would suggest using normal TCP calculation, i.e. that you dimension
the txqueue size after the "BDP rule": bandwidth x delay product. Using
your outbound numbers this is:
1,000,000 x 0.1 = 100,000 bits = 12,500 bytes. Assuming 1500 byte
packets, this is txqueue = 12,500 / 1500 = 8.3 packets. So you could try
txqueue size in the region 8-10, this should prevent queue starvation at
packet drop events caused by traffic overload, even if you only have one
single TCP flow."

[ CALCULATION ]

In my case with HTB and FQ_CODEL...
The latency is small 89ms (~0.1s)
DownLink ~4Mbps
MTU 1.500
( Is byte/packets equal to MTU value? )

4.000.000 (bits/s) x 0.1 (s) = 400,000 bits
400,000 (bits) ./. 8 = 50.000 (bytes)
50.000 (bytes) ./. 1.500 (byte/packets) = 33.3 packets

So a TXQUEUELEN of 33-35 seems to be OK?

root# ip link ls up
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
5: ppp0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1500 qdisc
pfifo_fast state UNKNOWN mode DEFAULT qlen 3
    link/ppp

root# ip link set txqueuelen 35 dev ppp0

root# ip link ls up dev ppp0
5: ppp0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1500 qdisc
pfifo_fast state UNKNOWN mode DEFAULT qlen 35
    link/ppp

I am testing here with SpeedOf.me.
Below some values with qlen=default=3.
[2] and [3] at my castle with recent (net-next + ppp-rtnetlink).
[1] at my brother's castle with an older kernel.
( D == DownLink and U == UpLink. )

SpeedOf.me results:

[1] TUE-TD: D:6.78 || U:3.04 (2015-02-22, w/ run_htb-with-fq_codel_ppp0.sh)
http://speedof.me/show.php?img=150221065031-67247.png

[2] DUS-SD: D:4.16 || U:2.18 (2015-12-31)
http://speedof.me/show.php?img=151231122405-1014.png

[3] DUS-SD: D:4.22 || U:1.63 (2015-12-31, w/ run_htb-with-fq_codel_ppp0.sh)
http://speedof.me/show.php?img=151231122154-8498.png

Suggestions?

Thanks.

- Sedat -

[1] http://www.spinics.net/lists/linux-ppp/msg00509.html
[2] http://speedof.me/

View attachment "dmesg_4.4.0-rc7-4-iniza-small.txt" of type "text/plain" (55636 bytes)

Download attachment "config-4.4.0-rc7-4-iniza-small" of type "application/octet-stream" (132261 bytes)

Download attachment "run_htb-with-fq_codel_ppp0.sh" of type "application/x-sh" (1481 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ