Transparent proxy not working after upgrade to 2.2

marcelloc

Just to be sure, Are you selecting interfaces on squid general tab?

I'm asking because other users had this "issue".

Brandhor

Just to be sure, Are you selecting interfaces on squid general tab?

I'm asking because other users had this "issue".

yeah I had that issue as well after upgrading the other day but I fixed it soon after

Brandhor

I tried to load a config.xml that looks as much as possible like a fresh installation and also did a checksum of all files and compared them but I don't think there was anything wrong with them

the only clue I have is the CLOSED:SYN_SENT state but it's not much

TieT

That means, you have send a sync request but no one answers (connection is closed)

Did you set a virtual IP on the WAN or LAN ?

Brandhor

@TieT:

That means, you have send a sync request but no one answers (connection is closed)

Did you set a virtual IP on the WAN or LAN ?

no

Koenig

I have this issue as well.

Foud this in the log:
php-fpm[83033]: /pkg_edit.php: The command '/usr/pbi/squid-i386/sbin/squid -k shutdown' returned exit code '1', the output was 'Shared object "libz.so.5" not found, required by "libmd5.so.1"'

marcelloc

@Koenig:

Foud this in the log:
php-fpm[83033]: /pkg_edit.php: The command '/usr/pbi/squid-i386/sbin/squid -k shutdown' returned exit code '1', the output was 'Shared object "libz.so.5" not found, required by "libmd5.so.1"'

Is this setup a migration from 2.1 or any 2.2RC version?

Not getting it on amd64. :-\

doktornotor

@Koenig:

Foud this in the log:
php-fpm[83033]: /pkg_edit.php: The command '/usr/pbi/squid-i386/sbin/squid -k shutdown' returned exit code '1', the output was 'Shared object "libz.so.5" not found, required by "libmd5.so.1"'

That thing gets removed. https://github.com/pfsense/pfsense/blob/master/etc/pfSense.obsoletedfiles


$ ls -l /usr/lib/libz.so*
lrwxr-xr-x  1 root  wheel  14 Jan 24 06:00 /usr/lib/libz.so -> /lib/libz.so.5

$ ls -l /lib/libz.so.5
ls: /lib/libz.so.5: No such file or directory

$ ls -l /lib/libz.so.6
-r--r--r--  1 root  wheel  79180 Jan 22 22:07 /lib/libz.so.6

This clearly is not done properly on FULL installs, as seen from the dangling dead symlinks.

Cf. that with nanobsd:


$ ls -l /usr/lib/libz.so*
lrwxr-xr-x  1 root  wheel  14 Jan 22 22:07 /usr/lib/libz.so -> /lib/libz.so.6

$ ls -l /lib/libz.so.6
-r--r--r--  1 root  wheel  79180 Jan 22 22:07 /lib/libz.so.6

@marcelloc:

Is this setup a migration from 2.1 or any 2.2RC version?
Not getting it on amd64. :-\

Apparently an upgrade process bug on full installs, rendering the updated library pretty much useless as it cannot be found. Just a few examples:


/usr/lib/libcrypto.so -> /lib/libcrypto.so.6
$ ls -l /lib/libcrypto.so.6
ls: /lib/libcrypto.so.6: No such file or directory

/usr/lib/libmd.so -> /lib/libmd.so.5
$ ls -l  /lib/libmd.so.5
ls: /lib/libmd.so.5: No such file or directory

$ ls -l /usr/lib/libusb.so*
lrwxr-xr-x  1 root  wheel     11 Jan 24 06:00 /usr/lib/libusb.so -> libusb.so.2
-r--r--r--  1 root  wheel  65068 Jan 22 22:07 /usr/lib/libusb.so.3

/usr/lib/libufs.so -> /lib/libufs.so.5
$ ls -l /lib/libufs.so.5
ls: /lib/libufs.so.5: No such file or directory

Badly, badly broken. >:( => https://redmine.pfsense.org/issues/4328

TieT

I think this can be resolved by creating a link between those libs.
and restarting squid.

ln -s /lib/libz.so.5 /usr/lib/libz.so.6

(if the file is in /lib of course)
You can do a search to make sure


su
find / -name libz.so.*

I had the same problem with squid and snort a while back
let me know what happens ;)

doktornotor

@TieT:

I think this can be resolved by creating a link between those libs.
and restarting squid.
ln /lib/libz.so.6 /lib/libz.so.5

Uh no, this is completely wrong "fix".

TieT

Uh no, this is completely wrong "fix".

So you're saying it won't work untill 2.2.1 is rolled out ?

doktornotor

You are fixing the wrong thing! That .5 vs .6 actually has a meaning - the new version of the library does not have a backward-compatible ABI! If you want to fix something, then fix the dead /usr/lib/*.so symlinks so that they point to proper file. Don't create "compatibility" symlinks pretending a deliberately removed outdated ABI library is still there.

marcelloc

Renato is going to rebuild squid pbi under i386, this may fix most issues.

TieT

Yes you're right, but do you expect users to have a non functional box for god knows how long.

like i said, its a temp fix that works for me.
when squid is rebuild and functioning again, I will undertake the needed steps to make it point to the right libs.

On the other hand, disable squid and wait for an update is also a solution…
But where is the fun in that :P

And yes, I know the .5 - .6 has a meaning, thank you for being so condescending.
also i made a typo, it should be```
ln -s /lib/libz.so.6 /usr/lib/libz.so

and i'm guessing squid points to /usr/lib/ to find the files ?

marcelloc

@TieT:

Yes you're right, but do you expect users to have a non functional box for god knows how long.

Is this question for me? ???

TieT

Is this question for me? ???

No ;)
Its was an answer to doktornotor

john880

Hello,

for me after the 2.1.5 to 2.2 version, the captive portal not worked.

I added an new Rules in Firewall to accept 8002 port and it's works.

marcelloc

@john880:

for me after the 2.1.5 to 2.2 version, the captive portal not worked.

Captive portal itself(better start a new thread) or Squid integration with captive portal?

Fredbovy

Hi,
My name is fred and I am sorry if I write in your post but but I did not find how to start a discussion
I have a problem with pfsense installation script was ok but now I cannot access to the box anymore. What can I do? Is it possible to reset the box?

You can write me directly fredbovyeirl@gmail.com

The in advance and sorry again
Fred

Steve Evans

@Brandhor:

although it's not a squid issue, I've tried to use netcat to listen to the same port as squid and I don't receive any traffic

this is what I get with pfctl while trying to connect to google.com from a client(192.168.1.9)
rdr on fxp0 inet proto tcp from any to ! (fxp0) port = http -> 127.0.0.1 port 3128
pass in quick on fxp0 proto tcp from any to ! (fxp0) port = 3128 flags S/SA keep state
fxp0 tcp 127.0.0.1:3128 (173.194.40.3:80) <- 192.168.1.9:55613       CLOSED:SYN_SENT

Hi Brandhor,

Have you got anywhere with this?

I started a similar thread pfSense 2.2: Squid 3.4.10_2 pkg 0.2.6 redirection not working in transparent mode before spotting you had the same issue.

You'll see below that I did exactly the same as you (great minds think alike) and came to the same conclusion.

@Steve:

To prove the point that the redirected transparent connection is not getting through to squid, I stopped the squid process and then ran the following on the pfSense console.
: nc -l 3128
I then telnetted to pfsense from my laptop and a connection was established and characters passed.
$ telnet 10.5.1.1 3128
Trying 10.5.1.1...
Connected to pfsense.scevans.com.
Escape character is '^]'.
hello
: nc -Dl 3128
hello
I repeated trying to telnet to port 80 on news.bbc.co.uk which should have been directed by the redirect, and no connection was made.

Steve

It was later in this thread that Marcelloc commented on the compile settings I'd got from squid.

@marcelloc:

@Steve:

Now here's an oddity.

This is one of pbi behaviors. same binary, different folder, different results. (Imagine get all it working together :))

And here is the confirmation that pbi build on i386 is outdated
'--disable-ipf-transparent' '--disable-ipfw-transparent'
Go to amd64 and it will work :)

Thanks for your feedback

Seeing that there was a firewall related compile time option I decided to wait for the squid update, but alas when that came the behaviour was unchanged.

Trying to understand what was going on I started another thread Order of rule application for redirects to which there has been no response.

Keen to hear where you are with this issue.

Thanks,

Steve