PPTP WAN (still) issue (MTU and fragmentation) going from 1.2.3 to 2.0RC3

doctor strangelove · Jul 15, 2011, 5:39 PM

Just tried both the MSS clamping (setting it to e.g. 1400) and removing the multilink (didn't seem to be on by default b.t.w. - only em1, which is my wan, was highlighted).

Unfortunately this didn't seem to work. I'll try again over the weekend with a fresh(er) install

rancor · Jul 16, 2011, 12:56 PM

@ermal:

Actually some 'clever' guy enabled multilink by default on pptp links.
Can you please confirm that removing multilink from interfaces->ppp, on your connection, fixes the issue?

I got the same problem connecting my Android to my LAN via PPTP but after I changed MTU to 1436 (WAN) it started to work. All data is routed from my Android via my LAN to Internet.

I does not how ever any PPP under Interfaces (or PPTP). Should I really have that after I created a PPTP VPN server?

Sylhouette · Jul 21, 2011, 11:41 AM

With the snapshot of today, still the same issue.

I can not find the multilink option under interfaces ppp !

regards,
Johan

Sylhouette · Aug 16, 2011, 12:32 PM

Downloaded the lataest 2.0 RC update file today.

But still the same issue, MTU settings 1436, 1400 but no web page will load, or some partitialy and some do not.
Ping to the sites work.

these i see a lot in the firewall log file
block
Aug 16 14:28:50 vr1 10.0.0.138 10.0.0.100 GRE
block
Aug 16 14:28:50 vr1 10.0.0.138 10.0.0.100 GRE

where 10.0.0.138 is the adsl modem, and 10.0.0.100 is the WAN interface.

If i use the old 1.2.3 pfsense all is working well

Thuersday i will reboot it to factory defaults and try if i can make it work from an fresh start.

regards,
Johan

Sylhouette · Aug 25, 2011, 9:01 AM

Well using a new installed system does not work either.

same logs 10.0.0.138 to 10.0.0.100 GRE dropped on vr1 (it is a soekriss 5501 running the i386 nanoimage).

This is getting a little frustrating.

I can use whatever verion of pfsense before2.0 and it works.
I can use any version of monowall and it works.
I can put my windows pc behind it install the nessasy pptp tools and it works
But pfsense 2.0 does not cope with it.

We have some branch offices who use this, all with pfsense 1.2.3
Our fiber is is also on pfsense 1.2.3, but i am getting a little reluctent to use version 2.

I hope someone will look at this, because for us this is a showstopper.

regards
Johan Hendriks

Sylhouette · Aug 25, 2011, 1:17 PM

This is the output of /tmp/rules.debug
It is set to factory defaults, just to try.

#System aliases

loopback = "{ lo0 }"
WAN = "{ pptp1 }"
LAN = "{ vr0 }"

#SSH Lockout Table
table <sshlockout> persist
table <webconfiguratorlockout> persist
#pfSnortSam tables
table <snort2c>

table <virusprot>

# User Aliases 

# Gateways
GWWAN = " route-to ( pptp1 195.190.242.11 ) "

set loginterface vr0
set optimization normal
set limit states 48000
set limit src-nodes 48000

set skip on pfsync0

scrub in on $WAN all    fragment reassemble
scrub in on $LAN all    fragment reassemble

nat-anchor "natearly/*"
nat-anchor "natrules/*"

# Outbound NAT rules

# Subnets to NAT 
tonatsubnets	= "{ 192.168.1.0/24 127.0.0.0/8  }"
nat on $WAN  from $tonatsubnets port 500 to any port 500 -> 213.84.84.84/32 port 500  
nat on $WAN  from $tonatsubnets to any -> 213.84.84.84/32 port 1024:65535  

# Load balancing anchor
rdr-anchor "relayd/*"
# TFTP proxy
rdr-anchor "tftp-proxy/*"
table <direct_networks> { 213.84.84.84/32 192.168.1.0/24 }
# UPnPd rdr anchor
rdr-anchor "miniupnpd"

anchor "relayd/*"
#---------------------------------------------------------------------------
# default deny rules
#---------------------------------------------------------------------------
block in log all label "Default deny rule"
block out log all label "Default deny rule"

# We use the mighty pf, we cannot be fooled.
block quick proto { tcp, udp } from any port = 0 to any
block quick proto { tcp, udp } from any to any port = 0

# Block all IPv6
block in quick inet6 all
block out quick inet6 all

# pfSnortSam
block quick from <snort2c> to any label "Block snort2c hosts"
block quick from any to <snort2c> label "Block snort2c hosts"
block quick from <pfsnortsamout> to any label "Block pfSnortSamOut hosts"
block quick from any to <pfsnortsamin> label "Block pfSnortSamIn hosts"

# SSH lockout
block in log quick proto tcp from <sshlockout> to any port 22 label "sshlockout"

# webConfigurator lockout
block in log quick proto tcp from <webconfiguratorlockout> to any port 443 label "webConfiguratorlockout"
block in quick from <virusprot> to any label "virusprot overload table"
antispoof for pptp1
# allow PPTP client
pass in on $WAN proto tcp from any to any port = 1723 flags S/SA modulate state label "allow PPTP client on WAN"
pass in on $WAN proto gre from any to any keep state label "allow PPTP client on WAN"
antispoof for vr0

# loopback
pass in on $loopback all label "pass loopback"
pass out on $loopback all label "pass loopback"
# let out anything from the firewall host itself and decrypted IPsec traffic
pass out all keep state allow-opts label "let out anything from firewall host itself"
pass out route-to ( pptp1 195.190.242.11 ) from 213.84.84.84 to !213.84.84.84/32 keep state allow-opts label "let out anything from firewall host itself"
# make sure the user cannot lock himself out of the webConfigurator or SSH
pass in quick on vr0 proto tcp from any to (vr0) port { 80 443  22 } keep state label "anti-lockout rule"

# User-defined rules follow

anchor "userrules/*"
pass  in  quick  on $WAN reply-to ( pptp1 195.190.242.11 )  from any to any keep state  label "USER_RULE"
pass  in  quick  on $LAN  from 192.168.1.0/24 to any keep state  label "USER_RULE: Default allow LAN to any rule"

# VPN Rules
anchor "tftp-proxy/*"</virusprot></webconfiguratorlockout></sshlockout></pfsnortsamin></pfsnortsamout></snort2c></snort2c></direct_networks></virusprot></snort2c></webconfiguratorlockout></sshlockout>

This is /tmp.rules.debug on the working 1.2.3 system (relevant part for as far i know)


# System Aliases 
loopback = "{ lo0 }"
lan = "{ vr0  }"
ng0 = "{ vr1 ng0 }" 
wan = "{ vr1 ng0  }"
enc0 = "{ enc0 }"
pptp = "{ ng1 ng2 ng3 ng4 ng5 ng6 ng7 ng8 ng9 ng10 ng11 ng12 ng13 ng14 ng15 ng16 }" 
WLAN = "{ vr2 }"
# User Aliases 

set loginterface vr1
set loginterface vr0
set loginterface vr2
set optimization normal

set skip on pfsync0
scrub all random-id  fragment reassemble

nat-anchor "pftpx/*"
nat-anchor "natearly/*"
nat-anchor "natrules/*"
# FTP proxy
rdr-anchor "pftpx/*"

# Outbound NAT rules
nat on $ng0 from 192.168.1.0/24 port 500 to any port 500 -> (ng0) port 500
nat on $ng0 from 192.168.1.0/24 port 5060 to any port 5060 -> (ng0) port 5060
nat on $ng0 from 192.168.1.0/24 to any -> (ng0) port 1024:65535
nat on $ng0 from 192.168.1.208/28 port 500 to any port 500 -> (ng0) port 500
nat on $ng0 from 192.168.1.208/28 port 5060 to any port 5060 -> (ng0) port 5060
nat on $ng0 from 192.168.1.208/28 to any -> (ng0) port 1024:65535

#SSH Lockout Table
table <sshlockout>persist

# Load balancing anchor - slbd updates
rdr-anchor "slb"

# FTP Proxy/helper
table <vpns>{   }
no rdr on vr0 proto tcp from any to <vpns>port 21
rdr on vr0 proto tcp from any to any port 21 -> 127.0.0.1 port 8021
no rdr on vr2 proto tcp from any to <vpns>port 21
rdr on vr2 proto tcp from any to any port 21 -> 127.0.0.1 port 8022

# NAT Inbound Redirects

# IMSpector rdr anchor
rdr-anchor "imspector"
# UPnPd rdr anchor
rdr-anchor "miniupnpd"

anchor "ftpsesame/*" 
anchor "firewallrules"

# We use the mighty pf, we cannot be fooled.
block quick proto { tcp, udp } from any port = 0 to any
block quick proto { tcp, udp } from any to any port = 0

# snort2c
table <snort2c>persist
block quick from <snort2c>to any label "Block snort2c hosts"
block quick from any to <snort2c>label "Block snort2c hosts"
# Block all IPv6
block in quick inet6 all
block out quick inet6 all
# loopback
anchor "loopback"
pass in quick on $loopback all label "pass loopback"
pass out quick on $loopback all label "pass loopback"

# package manager early specific hook
anchor "packageearly"

# carp
anchor "carp"

# permit wan interface to ping out (ping_hosts.sh)
pass quick proto icmp from 213.84.84.84 to any keep state

# NAT Reflection rules

# allow PPTP client
anchor "pptpclient"
pass in quick on $wan proto gre from any to any modulate state label "allow PPTP client"
pass in quick on $wan proto gre from any to any modulate state label "allow PPTP client"
pass in quick on $wan proto tcp from any port = 1723 to any flags S/SA modulate state label "allow PPTP client"
pass in quick on $wan proto tcp from any to any port = 1723 flags S/SA modulate state label "allow PPTP client"
block in log quick on $wan proto udp from any port = 67 to 192.168.1.0/24 port = 68 label "block dhcp client out wan"

# LAN/OPT spoof check (needs to be after DHCP because of broadcast addresses)
antispoof for vr0
antispoof for vr2

anchor "spoofing"
# Support for allow limiting of TCP connections by establishment rate
anchor "limitingesr"
table <virusprot>block in quick from <virusprot>to any label "virusprot overload table"

# block bogon networks
# http://www.cymru.com/Documents/bogon-bn-nonagg.txt
anchor "wanbogons"
table <bogons>persist file "/etc/bogons"
block in log quick on $wan from <bogons>to any label "block bogon networks from wan"

# let out anything from the firewall host itself and decrypted IPsec traffic
pass out quick on $lan proto icmp keep state label "let out anything from firewall host itself"
pass out quick on $wan proto icmp keep state label "let out anything from firewall host itself"

# tcp.closed 5 is a workaround for load balancing, squid and a few other issues.
# ticket (FEN-857512) in centipede tracker.
pass out quick on ng0 all keep state ( tcp.closed 5 ) label "let out anything from firewall host itself"
# pass traffic from firewall -> out
anchor "firewallout"
pass out quick on vr1 all keep state label "let out anything from firewall host itself"
pass out quick on vr0 all keep state label "let out anything from firewall host itself"
pass out quick on vr2 all keep state label "let out anything from firewall host itself"
pass out quick on $pptp all keep state label "let out anything from firewall host itself pptp"
pass out quick on $enc0 keep state label "IPSEC internal host to host"

# let out anything from the firewall host itself and decrypted IPsec traffic
pass out quick on vr2 proto icmp keep state ( tcp.closed 5 ) label "let out anything from firewall host itself"
pass out quick on $WLAN all keep state ( tcp.closed 5 ) label "let out anything from firewall host itself"

# make sure the user cannot lock himself out of the webGUI or SSH
anchor "anti-lockout"
pass in quick on vr0 from any to 192.168.1.250 keep state label "anti-lockout web rule"

# PPTPd rules
anchor "pptp"
pass in quick on $wan proto gre from any to 213.84.84.84 keep state label "allow gre pptpd"
pass in quick on $wan proto tcp from any to 213.84.84.84 port = 1723 modulate state label "allow pptpd 213.84.84.186"

# SSH lockout
block in log quick proto tcp from <sshlockout>to any port 22 label "sshlockout"

anchor "ftpproxy"
anchor "pftpx/*"

# IMSpector
anchor "imspector"

# uPnPd
anchor "miniupnpd"

#---------------------------------------------------------------------------
# default deny rules
#---------------------------------------------------------------------------
block in log quick all label "Default deny rule"
block out log quick all label "Default deny rule"</sshlockout></bogons></bogons></virusprot></virusprot></snort2c></snort2c></snort2c></vpns></vpns></vpns></sshlockout>

In both files i see the GRE protocol pass on the interface vr1. port 1723 (it is an soekriss 5501 board)
In my logs on the 2.0 i see that they are dropped from the vr1 interface.

The vr1 interface is not shown in my rules.debug file on the 2.0 machine but is on the 1.2.3 machine

On the working machine i have wan = "{ vr1 nr0}"
On the 2.0 i only have wan = "{ pptp0 }"

So the pass rules as far as i understand for the GRE protocol do not get set for the vr1 interface where i see all the drops in my log file.


# allow PPTP client
pass in on $WAN proto tcp from any to any port = 1723 flags S/SA modulate state label "allow PPTP client on WAN"
pass in on $WAN proto gre from any to any keep state label "allow PPTP client on WAN"

Could it be that i need to set wan = "{ vr1 pptp1 }" ?

regards
Johan Hendriks

mseiwald · Sep 14, 2011, 12:57 PM

I have exactly the same problem with 2.0RC3. It seems that MSS Clamping is somehow broken with PPTP in 2.0.

Sylhouette · Sep 15, 2011, 1:09 PM

Well i got it working, but it wil not survive a reboot.

what i did was the following

I did go to diagnostics, then edit file.
I opened /tmp/rules.debug
i put in the actual interface between the brackets of the wan interface.
wan = { pptp1 } and i changed it to wan = { vr1 pptp1}
Then i did go to diagnostics and then command
In the command box i did do pfctl -f /tmp/rules.debug

Then after hitting F5 in the browser which could not load any page, it all started to work, the pages were loaded.
To make sure it was this, i opened the traffic box and started to download a iso image from an ftp server, and the graph did flew up.

So the reason, it is not working in my case is that the vr1 interface does drop the GRE packages.
And the pf rules actually do let them pass, but the interface is not added to the wan.

regards,
Johan

mseiwald · Sep 18, 2011, 8:38 AM

Works, thanks!

edit:

I would like to make this survive a reboot with this script:


sed 's/{ pptp0 }/{ vr1 pptp0 }/g' /tmp/rules.debug > /tmp/rules.debug.new
mv /tmp/rules.debug.new /tmp/rules.debug
pfctl -f /tmp/rules.debug

I put it under /usr/local/etc/rc.d/fix_pptp like suggested in this thread but it is not executed at startup. Also pfsense does not automatically establish the PPTP-connection at startup. (I don't have Dial-On-Demand enabled)

Regards,

Michael

edit2:

Due to the connection being to unstable (I even had the above script in a cronjob every minute) I have downgraded to 1.2.3.

doctor strangelove · Sep 18, 2011, 10:01 AM

Hi,

Noticed 2.0-final was released yesterday so I couldn't resist trying it out to see if the pptp issue was resolved. Did a clean install just now but alas, the situation remains unchanged.

After that, decided to give Sylhouette's fix a try (for me it was adding em1 to pptp0) and this fixed the problem for me on the 2.0-final release!

Can't say anything yet about stability; I'll leave this running for a while and if it's stable, I'll try to get it reboot-proof.

rgds,
A.

Sylhouette · Sep 19, 2011, 8:20 AM

I think a lot of people get bitten by this.

Every firewall solution i tryed, work with the pptp solution.
monowall, the older pfsense, smoothwall, and sonicwall are the once i tried.
All without problem.

i need it to survive a reboot, we have a very poor power grid on one location, and power fails sometimes for more than 4 to 5 hours.
And this happens more than once per month.

Well we should see what we can do.
For now i leave these boxes at 1.2.3 because it works and works well.

regards.
Johan Hendriks

doctor strangelove · Sep 19, 2011, 9:55 AM

Hi,

Agree that it's a bit peculiar that something like pptp would be broken. Maybe pptp isn't all that popular anymore, don't know.

Your fix remained stable since I applied it. It's most probably not that hard to make it reboot proof by using some <shellcmd>statements in the config.xml. However, I guess (but haven't verified) that pfctl gets called whenever the ruleset is changed which would basically undo the commands issued from the config.xml (and break the pptp again)

The only reboot / filter-reload proof way to handle this seems to be to modify one (or multiple) scripts in the /etc (or /etc/rc.d) directories. Wouldn't want to do that on a production system since I don't know enough about the details of the inner workings of pfsense to feel confident that such hacking wouldn't result in undesirable behavior..

So, I'm (unfortunately) staying with 1.2.3. as well. Hope somebody can pick this issue up and raise a bug against it so maybe it can get resolved in an upcoming release..

b.r.
A.</shellcmd>

Sylhouette · Sep 20, 2011, 11:48 AM

Well i made it survive a reboot also, but this is an UGLY way.

go to diagnostics then edit file
open the file

/etc/inc/filter.inc

then around line 2221 you find the following.


# allow PPTP client
pass in on \${$oc['descr']} proto tcp from any to any port = 1723 flags S/SA modulate state label "allow PPTP client on {$oc['descr']}"
pass in on \${$oc['descr']} proto gre from any to any keep state label "allow PPTP client on {$oc['descr']}"

In my case my actual WAN interface is vr1 so i put a allow rule there on vr1.
Then it shows the following.


#allow PPTP client
pass in on \${$oc['descr']} proto tcp from any to any port = 1723 flags S/SA modulate state label "allow PPTP client on {$oc['descr']}"
pass in on \${$oc['descr']} proto gre from any to any keep state label "allow PPTP client on {$oc['descr']}"
pass in on vr1 proto gre from any to any keep state label "allow PPTP client on {$oc['descr']}"

This way it stays active after a reboot.
NOTE THIS IS NOT THE WAY TO DO IT..

But it works this way ….
And it seems we are the only one who have this.
On the other hand, it is saver than to set the wan value to vr1 and pptp0, because this way only GRE is allowed and in the rest of the script wan is just pptp0 like before.

It does not survive an upgrade however. :D

Gr
Johan Hendriks

mseiwald · Sep 21, 2011, 4:11 PM

Does anyone know if the developers are working on this issue or even are aware of it? Should someone (me?) create a ticket in redmine?

Sylhouette · Sep 22, 2011, 2:12 PM

I opened a thread on the mailling list.

When the 2.0 anouncement came, i repleyed on that by asking if this was resolved.

The following anwer came

That's not a general issue, and haven't heard that from anyone else,
not sure what you're seeing.

I replied on that by explaining what i saw and how in my case i got that fixed.
I also reffered to this thread on the forum.

Then to my surprise i almost got a ban from the mailing list, this was the answer i got on my relpy.

Stop here, you're hijacking a thread. Bump your own thread if you want
to continue.

I replied again with a sorry that it was not my intention to hijajack a thread. ( i really was not )
And that i already had a threat on the mailling list about this issue and gave the link to the thread on the mailling list.

Never heard anything after that.
I did not send an e-mail to the mailling list after that so maybe i am already banned :(, i hope not, because i like to help out sometimes, and i learn a lot about PFSense through the mailling list.

So i really do not know the status.

regards,
Johan Hendriks

doctor strangelove · Sep 25, 2011, 3:23 PM

Hi,

Opening a ticket is probably the best way forward since several people clearly have the same reproducible problem with PPTP (some might call that a bug ;)) . I don't know how to get from a discussion on the forum to an actual ticket in pfsense's bugtracker - mseiwald, if you are capable of raising a ticket, yes please!

Probably a good idea to include Sylhouette's two fixes in the bug description as well.

rgds,
A.

doctor strangelove · Dec 21, 2011, 10:55 AM

Hi all,

Just noticed 2.0.1 is out! Before I give it a try, I was wondering if anyone knows if the PPTP WAN problem that existed since 2.0RC1 has been properly fixed yet. Doesn't seem to be (explicitly) in the list of changes in the release notes but who knows..

br.
A.

bill_mcgonigle · Feb 21, 2012, 8:34 PM

I'm seeing an MTU issue after upgrading from 1.2.3 to 2.0. I 'fixed' this by editing

/var/etc/pptp-vpn/mpd.conf

and adding:

set link mtu 1396

Then, killing the mpd process and restarting it:

/usr/local/sbin/mpd4 -b -d /var/etc/pptp-vpn -p /var/run/pptp-vpn.pid -s pptps pptps

That's not a pfSense 'solution' but perhaps others here could try it to see if our problems are the same?

This wasn't necessary before, so I'm wondering if defaults are different, path MTU discovery is somehow broken, etc. BTW, I read that XP's pptp client requires 1396, so that's what I set for a compatibility floor.