TCAM and Access Lists

Hi
I am getting myself confused with the TCAM tables and access-lists. From reading the switching book I can see that the access-lists are compiled into the TCAM. What is confusing me is how a switch does a lookup in parallel when the access lists are written in a specific order. You can't parallel something which has to be done in sequence.
Am I just misreading this part and the author means that each statement is looked up in a parallel table lookup, in serial order?
Also how do we achieve the parrallelism, is there a multiprocessor asic doing something weird and wonderful in there with parallel code like OCCAM? (showing my age there :) )

Hi Carl,
As i mentioned eariler, we could say CBAC is a kind of Access list.
Access list is just filtering some prefixes/ports. CBAC is the additional feature over ACL. We make use of the ACL in CBAC to inspects traffic that travels through the firewall to discover and manage state information for TCP and UDP sessions.
Just go thro. this link for both ACL and CBAC,
http://www.cisco.com/en/US/products/sw/secursw/ps1018/products_tech_note09186a00800a5b9a.shtml#cbac
Rgs,
Balaji. V

Similar Messages

2800 NAT & Access Lists

Hi Folks,
Hope I'm posting in the right section...
I have a 2811 router that I want to configure the F/E 0/0 & 0/1 to be able to communicate.
FE 0/0 is configured with 192.168.16.1/24
FE 0/1 is configured with 10.10.10.10/24
     Physically attached to this port is going to be a wireless router configured with 10.10.10.1.
My question is how to configure and setup the routes properly and do I need to configure NAT and Access Lists.
TIA.

thanks for the reply...
I'll try to fill in some blanks but here's where I'm at.
I'm working with a remote 2811 and a TP-Link Wireless router. The 2811 is configured as follows...
interface FastEthernet0/0
ip address 192.168.16.1 255.255.255.0
duplex auto
speed auto
no keepalive
interface FastEthernet0/1
description fdrvacctv subnet
ip address 10.10.10.10 255.255.255.0
ip access-group 1 in
duplex auto
speed auto
access-list 1 permit 192.168.1.0 0.0.0.255
access-list 1 permit 192.168.10.0 0.0.0.255
         access-list 1 permit 192.168.11.0 0.0.0.255
         access-list 1 permit 192.168.16.0 0.0.0.255
         access-list 1 permit 192.168.17.0 0.0.0.255
         access-list 1 permit 10.10.10.0 0.0.0.255
Attached to F/E 0/1 is my wireless router via one of it's LAN ports. The LAN side on the wireless router is set with a static IP of 10.10.10.1.
From my 2811 I can ping 10.10.10.1 or 10.10.10.5 (a pc attached to that wireless). I can not ping 10.10.10.1 from 192.168.16.1
2811#ping 10.10.10.1
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.10.10.1, timeout is 2 seconds:
Success rate is 80 percent (4/5), round-trip min/avg/max = 1/1/1 ms
#ping 10.10.10.1 source 192.168.16.1
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.10.10.1, timeout is 2 seconds:
Packet sent with a source address of 192.168.16.1
Success rate is 0 percent (0/5)
2811#
It's this scenario that leads me to belive my problem is with either a static route, access-list, or both. Perhaps I need ot NAT something?

Cisco ISE and WLC Access-List Design/Scalability

Hi,
I have a scenario whereby wireless clients are authenticated by the ISE and different ACLs are applied to it based on the rules on ISE. The problem I seems to be seeing is due to the limitation on the Cisco WLC which limit only 64 access-list entries. As the setup has only a few SVI/interfaces and multiple different access-lists are applied to the same interface base on the user groups; I was wondering if there may be a scalable design/approach whereby the access-list entries may scale beside creating a vlan for each user group and applying the access-list on the layer 3 interface instead? I have illustrated the setup below for reference:
User group 1 -- Apply ACL 1 --On Vlan 1
User group 2 -- Apply ACL 2 -- On Vlan 1
User group 3 -- Apply ACL 3 -- On Vlan 1
The problem is only seen for wireless users, it is not seen on wired users as the ACLs may be applied successfully without any limitation to the switches.
Any suggestion is appreciated.
Thanks.

Actually, you have limitations on the switch side as well. Lengthy ACLs can deplete the switch's TCAM resources. Take a look at this link:
http://www.cisco.com/c/en/us/support/docs/switches/catalyst-3750-series-switches/68461-high-cpu-utilization-cat3750.html
The new WLCs that are based on IOS XE and not the old Wireless/Aironet OS will provide the a better experience when it comes to such issues.
Overall, I see three ways to overcome your current issue:
1. Shrink the ACLs by making them less specific
2. Utilize the L3 interfaces on a L3 switch or FW and apply ACLs there
3. Use SGT/SGA
Hope this helps!
Thank you for rating helpful posts!

Vpn site to site and remote access , access lists

Hi all, we run remote access and site to site vpn on my asa, my question is Can I create an access list for the site to site tunnel, but still leave the remote access vpn to bypass the access list via the sysopt command, or if I turn this off will it affect both site to site and remote access vpn ?

If you turn off sysopt conn permit-vpn it will apply to both your site to site and remote access vpn...all ipsec traffic. You would have to use a vpn-filter for the site to site tunnel if you wanted to leave the sysopt in there.

I have a 3rd generation iPod Touch and just did the update to IOS 5. Now I can't connect to my Netgear wifi router. My iPhone connects fine along with all of my other laptops etc. I have the router set with WPA-PSK [TKIP] security and an access list.

I have a 3rd generation iPod Touch and just did the update to IOS 5. Now I can't connect to my Netgear wifi router. My iPhone connects fine along with all of my other laptops etc. I have the router set with WPA-PSK [TKIP] security and an access list. I've confirmed the mac address is included on that list and that the password is correct. Under choses netwrok I select the network and it just goes into a spin. I have tried removing the password and the access list settings and it still will not complete the connection to the router thus no internet access. The routers firmware is also up to date. This thing worked fine before this update and I've already tried to restore from backup. Any ideas or is the wifi nic bad in this thing with the new apple firmware update? Any fix?

Thanks Bob, I don't know why but it all of a sudden worked a few days later. It's a mystery but at least problem solved.

Access List and Conflict Resolution Problem!

My configuration for Allow and Deny is not allowing me to load images and CSS files through the gateway on a URLScraper channel.
I'm trying to figure out how to control access to resources using the Access List service, and I'm running into trouble. The Sun ONE Portal Server, Secure Remote Access 6.0 Administrator's Guide (Doc 816-6421-10) states:
Setting the Conflict Resolution Level
You can set the priority level for the dynamic attributes. If a user inherits multiple attribute templates, say from an organization and a role assignment, and there is a template conflict between the attributes in the two templates, the template with the highest priority is inherited. There are seven settings available ranging from Highest to Lowest.
See the Administration Guide, iPlanet Directory Server Access Management Edition for more details on conflict resolution.
Unfortunately the referenced Adminstration Guide for DSAME contains exactly 0 occurances of the word "conflict" in its 136 pages, so that reference was less than helpful. Chapter 17 of that document (Doc 816-5620-10) describes URL Policy Agent Attributes, which sheds some light on what the URL Deny and URL Allow settings mean. The key sentence is, "An empty Deny list will allow only those resources that are allowed by the Allow list."
So, I've set up my Access List services as follows:
o URL Deny is blank on all Access Lists
o URL Allow set as follows
---- isp
------- http://portal.acme.com/portal/* (company name changed to protect the guilty!)
---- acme.com organization
------- Conflict Resolution: Highest
------- http://portal.acme.com/portal/* (same as above)
---- Acme Customers Role - shared role for all Acme customers
------- Conflict Resolution: Medium
------- http://www.acme.com/*
------- http://support.acme.com/*
------- http://support2.acme.com/*
---- RoadRunner role - specific role for a specific customer
------- Conflict Resolution: Medium
------- http://roadrunnerinfo.acme.com/*
The Desktop services in each of the above two roles includes channels from the hosts in the URL Allow lists.
The behavior I'm seeing with this configuration is that the desktop channels include information from the scraped HTML, and the URLs are rewritten for the included images and CSS files and such. However, the gateway is denying access to the images referenced by the rewritten URL. That is, an image with a URL of https://portal.acme.com/http://roadrunnerinfo.acme.com/images/green.gif shows up as a broken image on the desktop. Attempting to access the URL to the image directly results in an "Access to this resource is denied !! Contact your administrator" error message.
If I set the conflict resolution on the acme.corp organization to Medium (or anything lower than the two role conflict resolution levels) results in the same error message as soon as the customer logs in (no desktop rendered). The same error occurs if I set the conflict resolution in the two roles to Highest (same as the top level organization), again with no desktop rendered on login.
If I put all the above referenced URLs in the acme.com organization Access List service, then I am successfully able to fetch all the resources (images, CSS, etc.) in the URLScraper HTML. Likewise if I put "*" in that Access List. However, this is less than ideal, as it would potentially allow other customers to view data that isn't theirs (Wile E. Coyote user should not be able to get to Road Runner data, and vice versa, and neither one of them should get at Acme private information!).
So, what am I doing wrong? Also, does anyone have any leads on where I can read up on how Access Lists and conflict resolution are supposed to work, since Sun neglected to include a valid reference in the Administrator's Guide, Portal Server 6.0 SRA?
Thanks!
-matt

Did you ever get anywhere with this. My experiments seem to inidicate that you cannot successfully combine Access and Deny directives, across roles or organizational defaults and a role.

ICMP Inspection and Extended Access-List

I need a little help clarifying the need for an Extended Access-list when ICMP Inspect is enabled on an ASA. From reading various documents such as the following (http://www.cisco.com/c/en/us/support/docs/security/pix-500-series-security-appliances/15246-31.html), I CAN allow ICMP through my ASA using an extended access-list or enabling ICMP Inspection in the Modular Policy Framework. Is that true? I only NEED an Extended Access-list or enable ICMP Inspection? I do not need both? Or is it best practice to do both?
What does the ASA do to a PING from a host on the inside interface (Security 100) to host on the outside interface (Security 0) when ICMP Inspection is enabled with the following commands:
policy-map global_policy
class inspection_default
inspect_icmp
However, the following commands are NOT placed on the inbound Extended Access-list of the outside interface:
access-list inbound permit icmp any any echo-reply
access-list inbound permit icmp any any source-quench
access-list inbound permit icmp any any unreachable
access-list inbound permit icmp any any time-exceeded
access-group inbound in interface outside
Will the PING complete?
Thank you,
T.J.

Hi, T.J.
If problem is still actual, I can answer you this question.
Let's see situation without ICMP inspection enabled:
The Cisco ASA will allow ICMP packets only in case if ACL entry exist on interface, where packet goes in. If we're speaking about ping, then ACL rules must allow packets in both directions.
In case with ICMP inspection, with ACL entry you should allow only request packets, replies will be allowed based on ICMP inspection created connection.
Speaking about your particular example with different security levels - with default ACL rule, that allow traffic from higher interface to lower - NO, you can do not enter that rules you described, and as you'll have successful ping.
If you deleted this rule and administrate allowed traffic manually, then YES, you must allow ICMP requests to have successful ping.
P.S. It's not a good practice to leave that default rule, which allow traffic from higher sec.lvl. to lower.

ASR 5000 access list for ssh and telnet

Dears,
how can we apply an access list for telnet and ssh on asr 5k ?
please advise if this is feasible.
thx.

Hello Joseph,
Sorry for the delay in response.
To control access to ASR5000 via telnet, other than configuring an ACL, there is a way to disable telnetd by configuring local context.
For example:
config
context local
no server telnetd
#exit
System Administration Guide of the relevant version will give you detailed information in this regard.
Here is the latest system admin guide (for SW version 17): http://www.cisco.com/c/dam/en/us/td/docs/wireless/asr_5000/17-0/PDF/17-ASR5000-Sys-Admin.pdf
You can find other guides here: http://www.cisco.com/c/en/us/support/wireless/asr-5000-series/products-installation-and-configuration-guides-list.html
Hope this helps..
Regards
Aneesh

RoleEntityACL|Role Access List | no values, guest and authenticated not shown

All,
I enabled RoleEntityACL from configuration manager. Role access list field shows up, but when I type **, there are no values at all. Not even guest and authenticated, OOTB values.
I added UseEntitySecurity=true, I am able to see add Users and Groups when i type ** in the input field.
Any pointers here?
Thanks
~

Srinath,
I need to see the guest and authenticated values by default after enabling the "RoleEntityACL". Am i missing something here. I have Roles text box enabled, but it is not giving any values even if type **, g or a.
If i get those values, i can go to configuration manager applet and then add more values.
However, I did all those u mentioned. Added a new role in ExternalRoleView, Published Schema and Schema Base. Restarted UCM server. But i see null results.
In General Configuration:
UseEntitySecurity=true
SpecialAuthGroups=TestGroup,Public
In Advanced Component Manager:
Enabled RoleEntityACL.
I am able to add users and groups(aliases) in the access control list at the folder level. but not any roles. Am i missing something here?

WAAS and SNMP access-list

I am using 4.1.1c(build b16), and testing restricting access to the SNMP MIBS. we are running inline with a separate interface for mgmt. (gi1/0). with snmp access-list defined and snmp-server access-list set.
snmp-server community public
snmp-server access-list SNMP
ip access-list standard SNMP
permit 10.10.10.2
when i walk the mib from 10.10.10.2, and then look at ACL, it doesn't show any access.
CM#sh ip access-list SNMP
Standard IP access list SNMP
1 permit 10.10.10.2
(implicit deny any: 0 matches)
total invocations: 0

To define an IP ACL from the CLI, you can use the ip access-list global configuration command, and to apply the IP ACL to an interface on the WAAS device, you can use the ip access-group interface configuration command. To configure the use of an IP ACL for SNMP, you can use the snmp-server access-list global configuration command. To specify an IP ACL that the WAE applies to the inbound WCCP GRE encapsulated traffic that it receives, you can use the wccp access-list global configuration command.

VLAN's, subinterface, access-lists and 3560 catalyst switch?

Hi,
How can I isolate VLAN 121 from all others?
I have a cisco 2811 router connected to a 3560 catalyst switch which has 5 VLAN's of which I need to protect IP traffic of 4 from 1.
The following VLANs configured on the switch:
VLAN 0 192.168.132.0 /24
VLAN 135 ..135.0 /24
VLAN 137 ..137.0 /24
VLAN 139 ..139.0.24 and lastly,
VLAN 121 192.168.121.0 /24 which I wish to isolate all IP from VLAN 0, 135, 137, and 139 but have internet out the 2811's other interface. Currently all VLAN's and routing are working perfectly.
I need some advice please. Here is my plan: to split the FA0/0 into FA0/0.1 for VLAN 121 using dot1q and apply an access-list to deny 192.168.121.0 to the FA0/0 interface. Since I'm essentially creating VLAN's with the router can or will that interfere with the Switch VLAN configuration? router on a stick vs. a Layer 4 Cisco 3560 Catalyst switch?
Thank you!

I will have to assume VLAN 0 is the native VLAN / default interface on the router? All VLANs are numbered native or not. Just ensure the VLAN numbering matches between the router and the trunking on the switch.
Yes, you could create a sub interface on the 2811 and use the router to route the VLAN. Apply an access list on the other interfaces to block access to the VLAN you want to protect. If you have routing enabled on the 3560 as well you would complicate the situation a bit more.
Please rate helpful posts! :-)

Clienless webvpn and reflexive access list firewall

I have a Cisco Router 3825 with WEBVPN server and Reflexive access list Firewall. All is well but when i try from outside to go to WEBVPN server and try trought WEBVPN site to open some web Site it dosen`t work. For example when i try to open yahoo.com, the log shows
"%SEC-6-IPACCESSLOGP: list ACL-FILTER-IN denied tcp 98.138.253.109(80) -> my_ip_address(45341), 1 packet [ACL_ERROR]"
98.138.253.109 is yahoo.com ip address
Can you give my advice how to solve this problem?

If you have WEBVPN, then you have the Security-image/license on your router. That means that you are not restricted to reflexive ACLs, you can use a "real" firewall-feature like CBAC or ZBF on that device.

ACE access-list and Passive FTP

Can servers sitting behind the ACE successfully ftp files if the following rules are in place?
access-list word line x extended permit tcp source destination eq 21
access-list word line y extended permit tcp source destination eq 20
With those lines I can establish an FTP session, but unable to transfer files.
With the following statement access-list word line x extended permit ip source destination, passive ftp works?
IS this because the ACE acl does not allow for stateful inspection of an FTP session?
Thank you

You are right lack of fixup/inspect is the reason for FTP connections to fail.
You need something in line with the following config
class-map match-all FTP-Traffic
2 match port tcp eq ftp
policy-map multi-match xyz
class FTP-Traffic
inspect ftp
Syed Iftekhar Ahmed

IOS XR deny ace not supported in access list

Hi everybody,
We´ve a 10G interface, this is a MPLS trunk between one ASR 9010 and a 7613, and the first thing that we do is through a policy-map TK-MPLS_TG we make a shape of 2G to the interface to the output:
interface TenGigE0/3/0/0
cdp
mtu 1568
service-policy output TK-MPLS_TG
ipv4 address 172.16.19.134 255.255.255.252
mpls
mtu 1568
policy-map TK-MPLS_TG
class class-default
service-policy TK-MPLS_EDGE-WAN
shape average 2000000000 bps
bandwidth 2000000 kbps
and we´ve the policy TK-MPLS_EDGE-WAN as a service-policy inside, this new policy help us to asign bandwidth percent to 5 class-map, wich in turn match with experimental values classified when they got in to the router:
class-map match-any W_RTP
match mpls experimental topmost 5
match dscp ef
end-class-map
class-map match-any W_EMAIL
match mpls experimental topmost 1
match dscp cs1
end-class-map
class-map match-any W_VIDEO
match mpls experimental topmost 4 3
match dscp cs3 cs4
end-class-map
class-map match-any W_DATOS-CR
match mpls experimental topmost 2
match dscp cs2
end-class-map
class-map match-any W_AVAIL
match mpls experimental topmost 0
match dscp default
end-class-map
policy-map TK-MPLS_EDGE-WAN
class W_RTP
bandwidth percent 5
class W_VIDEO
bandwidth percent 5
class W_DATOS-CR
bandwidth percent 30
class W_EMAIL
bandwidth percent 15
class W_AVAIL
bandwidth percent 2
class class-default
end-policy-map
what we want to do is to assign a especific bandwidth to the proxy to the output using the class W_AVAIL, the proxy is 150.2.1.100. We´ve an additional requirement, wich is not apply this "rate" to some networks we are going to list only 4 in the example, so what we did was a new policy-map with a new class-map and a new ACL :
ipv4 access-list PROXY-GIT-MEX
10 deny ipv4 host 150.2.1.100 10.15.142.0 0.0.0.255
20 deny ipv4 host 150.2.1.100 10.15.244.0 0.0.0.255
30 deny ipv4 host 150.2.1.100 10.18.52.0 0.0.0.127
40 deny ipv4 host 150.2.1.100 10.16.4.0 0.0.0.255
50 permit tcp host 150.2.1.100 any
60 permit tcp host 10.15.221.100 any
policy-map EDGE-MEX3-PXY
class C_PXY-GIT-MEX3
police rate 300 mbps
class class-default
end-policy-map
class-map match-any C_PXY-GIT-MEX3
match access-group ipv4 PROXY-GIT-MEX
end-class-map
we asign a policy rate of 300 mbps to the class inside the policy EDGE-MEX3-PXY and finally we put this new policy inside the class W_AVAIL of the policy TK-MPLS_EDGE-WAN
policy-map TK-MPLS_EDGE-WAN
class W_RTP
bandwidth percent 5
class W_VIDEO
bandwidth percent 5
class W_DATOS-CR
bandwidth percent 30
class W_EMAIL
bandwidth percent 15
class W_AVAIL
service-policy EDGE-MEX3-PXY
class class-default
end-policy-map
and we get this:
Wed Sep 17 18:35:36.537 UTC
% Failed to commit one or more configuration items during a pseudo-atomic operation. All changes made have been reverted. Please issue 'show configuration failed' from this session to view the errors
RP/0/RSP1/CPU0:ED_MEX_1(config-pmap-c)#show configuration failed
Wed Sep 17 18:35:49.662 UTC
!! SEMANTIC ERRORS: This configuration was rejected by
!! the system due to semantic errors. The individual
!! errors with each failed configuration command can be
!! found below.
!!% Deny ace not supported in access-list: InPlace Modify Error: Policy TK-MPLS_TG: 'km' detected the 'warning' condition 'Deny ace not supported in access-list'
end
Any kind of help is very appreciated.

That is correct, due to the way the class-matching is implemented in the TCAM, only permit statements in an ACL can be used for QOS class-matching based on ACL.
unfortunately, you'll need to redefine the policy class match in such a way that it takes the permit only.
if you have some traffic that you want to exclude you could do something like this:
access-list PERMIT-ME
1 permit
2 permit
3 permit
access-list DENY-me
!the exclude list
1 permit
2 permit
3 permit
policy-map X
class DENY-ME
<dont do anything> or set something rogue (like qos-group)
class PERMIT-ME
do here what you wanted to do as earlier.
eventhough the permit and deny may be overlapping in terms of match.
only the first class is matched here, DENY-ME.
cheers!
xander

UWC/CE 6.3 and Access Manager 7.1 SSO sometimes fails (seems like a bug)

PREAMBULA: I started writing this post thinking that our AM SSO setup was at fault in some step. As I was gathering data, checking the doc-links and config files and finally sniffed the servers for HTTP dialogs, I grew pretty sure there's a bug in UWC/CE, AM SDK or Web Server Policy Agent, whatever implements the AM SSO session checking.
In short, as written below, our "sunmail" server can POST a broken cookie to AM server, if the cookie originally contained a "plus" character. The "plus" is replaced by a "space", invalidating the session check. As we know, "+" is often used in URLs to "escape" the space character. Perhaps some URL cleanup routine backfired here. I have double-checked, it is not the reverse proxy on "psam" breaking things. It is "sunmail" (UWC/CE or Policy Agent, don't know for certain) supplying the broken request. On the few occasions when the AM cookie contains no "plus" characters, the SSO works like a charm (also checked by a sniffer). Whenever there is a "plus", it breaks.
Is there some known bug or workaround that matches this description?
Nevertheless, for completeness' sake I kept the description of our setup. Maybe it's at fault after all :)
We have an installation of JCS5 with the latest patches as of early July 2008. And as the subject implies, we have problems with AM SSO in UWC/CE web-interface. I have reported them before, then they seemed fixed (not occuring for several tests in a row), but as time has shown, something wrong is still there.
So I'll try to go into deeper detail now, as we've may have overlooked some nuance... Then again, as my sniffer research below shows, this may be an engine bug and these setup details are irrelevant.
Our setup is split into several Solaris 10 full-root zones hosted on several servers, some of the components are enroute to HA (perhaps we made some mistakes on this part of the way?)
So, we have the following software stack:
1) two MMR Directory Servers (DSEE 6.3 = DSEE 6.2 from JCS5 + 125278-07__DSEE_6.3__x86x64 + 125277-07__DSEE_6.3__x86_sol9 patches) working in zones on two different servers. Except for one time when a manually forced ZFS rollback corrupted one of the server instances, no problems here.
2) two zones with Directory Proxy Servers (6.3, exact versions as above) running at port 389 provide the clients with an illusion that they have a stable Directory Server, even if one of the actual servers is currently rebooting ;)
These DPS zones are hosted on two different servers as well and are primarily used by LDAP clients (JCS components) running in other zones on the same respective servers.
3) A zone with Sun Web Server 7.0U1 and Access Manager 7.1 (+ 126357-01__AM71_x86 patch) and Delegated Admin 6.4-4.01 (from JCS5 + 121582-18__COMMCLI64__x86 patch).
At the moment there is one such zone (named "cos-psam-01.domain.ru" in the logs below), but we expect(-ed) it to become two similar zones as per AM HA setup.
Zones listed in (1-3) use private IP numbers, they belong in our internal DMZ.
Zones listed in (4-5) below use public (routed) IP numbers, they belong in our external DMZ.
4) A zone with Sun Web Server 7.0U1 used primarily as a reverse-proxy server (optionally with a load-balancer libpassthrough.so plugin) successfully used for other hosted projects. One of its configurations now passes connections from an externally routed IP address published as "psam.domain.ru" to "cos-psam-01.domain.ru", per AM HA setup, so HTTP clients believe they work with an Access Manager instance. This zone has a backend interface with a private IP address to communicate with the actual AM instance.
In AM configuration (both LDAP and file-based) we have configured a site ID with the publicly known name and mentioned both names (psam and cos-psam-01) in organization's realm/dns aliases.
5) A zone with the rest of the Sun Java Communications Suite 5, as in Messaging Server 6.3 (6.3-6.03 64-bit: ci-5.0-1.03_solx86_x64__Messaging_Server_6.3-2 + patch 126480-09__MSG63__x86-64), UWC/CE 6.3 (from JCS5 + 122794-17__UWC63-4.01_core__x86), Instant Messaging 7.2 (from JCS5 + 118790-29__IM72__x86-1 + 118787-28__IM72__x86-2), Calendar Server 6.3 (from JCS5 + 121658-28__iCS63__x86). The web-components (UWC/CE, IM, /httpbind) are deployed in a Sun Web Server 7.0U1 as well.
This zone is named "sunmail.domain.ru" and has a routed IP address for direct external access to its servicess.
The AM SDK part is also patched (126357-01__AM71_x86); it points to the load-balancer name ("psam.domain.ru") as an actual AM server.
# imsimta version
Sun Java(tm) System Messaging Server 6.3-6.03 (built Mar 14 2008; 64bit)
libimta.so 6.3-6.03 (built 17:15:08, Mar 14 2008; 64bit)
SunOS sunmail 5.10 Generic_127112-07 i86pc i386 i86pc
While setting up this server set we tried to use AM SSO as the user login method, but it works unreliably.
"Unreliably" means that while most of the time entering a correct uid and password in Access Manager login page ("http://psam.domain.ru/amserver/UI/Login") does redirect a user back to "http://sunmail.domain.ru/uwc/auth" along with a new cookie, and the user is redirected again to his or her mailbox, sometimes the user receives the UWC/CE login page. Entering the same uid and password here does log him in, but it breaks the whole point of SSO and only increases the end-user routine required to log in :\
We have also seen the "missing mail tab" problem - if the users point the browser to any hostname different from "sunmail.domain.ru" (i.e. www.mail.domain.ru which is equivalent in DNS), they have only the Address book, Calendar and Options tabs; no webmail. So far this is resolved by Policy Agent forcing The One name of the server.
Here's the configuration we did specifically for AM SSO:
1) in AMConfig.properties of "sunmail" and "cos-psam-01" we set up
com.iplanet.am.cookie.encode=false
am.encryption.pwd=<the same value>
all hostname-related parameters point to "psam.domain.ru"
2) in AMConfig.properties of "cos-psam-01" a number of FQDN equivalence entries are added (so it does not redirect to a server hostname unknown to visitors):
com.sun.identity.server.fqdnMap[publicname-or-ip]=psam.domain.ru
com.sun.identity.server.fqdnMap[cos-psam-01.domain.ru]=cos-psam-01.domain.ru
3) in "msg.conf" on "sunmail" (entries added via configutil):
local.webmail.sso.amcookiename = iPlanetDirectoryPro
local.webmail.sso.amnamingurl = http://psam.domain.ru:80/amserver/namingservice
local.webmail.sso.singlesignoff = yes
local.webmail.sso.uwcenabled = 1
service.http.ipsecurity = no
(perhaps some more options are required? Looking for confirmation about: local.webmail.sso.uwclogouturl local.webmail.sso.uwccontexturi local.webmail.sso.uwchome service.http.allowadminproxy )
4) Configured Web Policy Agent for Sun Web Server, so that users without an AM session are required to get one. Set up per [http://msg.wikidoc.info/index.php/AM_redirection_using_Policy_Agent], except that com.sun.am.policy.agents.config.notenforced_list points to the many names our server can go known by.
5) Updated the logout URL in /opt/SUNWuwc/webmail/main.js:
--- main.js.orig        Sat Jan 26 07:52:09 2008
+++ main.js     Mon Jul 21 01:06:29 2008
@@ -667,7 +667,8 @@
function cleanup() {
   if(laurel)
-      top.window.location = getUWCHost() + "/base/UWCMain?op=logout"
+//      top.window.location = getUWCHost() + "/base/UWCMain?op=logout"
+      top.window.location = "http://sunmail.domain.ru:80/base/UWCMain?op=logout"
   else
       exec('logout', '', 'exit()')
@@ -1707,7 +1708,8 @@
   if(lg) {
         url = document.location.href
         url = url.substr(0,url.indexOf('webmail'))
-        uwcurl = url + 'base/UWCMain?op=logout'
+//      uwcurl = url + 'base/UWCMain?op=logout'
+        uwcurl = "http://sunmail.domain.ru:80/base/UWCMain?op=logout"
   exit()
}6) Calendar SSO - per docs...
According to ngrep sniffing,
1) the browser goes to "http://sunmail.domain.ru/uwc/auth" without any cookies
2) receives a redirect and goes to "http://psam.domain.ru/amserver/UI/Login?gotoOnFail=http://sunmail.domain.ru:80/uwc&goto=http%3A%2F%2Fsunmail.domain.ru%3A80%2Fuwc%2Fauth"; sends no cookies either.
3) The first response from the "psam" server (as redirected from "cos-psam-01") sets a few cookies while rendering the login page:
Set-cookie: JSESSIONID=7EF8F2810D2071CA03CFEAE9972735B2; Path=/
Set-cookie: AMAuthCookie=AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#; Domain=.domain.ru; Path=/
Set-cookie: amlbcookie=02; Domain=.domain.ru; Path=/
4) The browser requests the login page resources (javascripts, images, etc) using these cookies, as in this header line:
Cookie: JSESSIONID=7EF8F2810D2071CA03CFEAE9972735B2; AMAuthCookie=AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#; amlbcookie=02
5) The browser POSTs the login request to "/amserver/UI/Login" and receives a redirection to http://sunmail.domain.ru:80/uwc/auth
Set-cookie: iPlanetDirectoryPro=AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#; Domain=.domain.ru; Path=/
Set-cookie: AMAuthCookie=LOGOUT; Domain=.domain.ru; Expires=Thu, 01-Jan-1970 00:00:10 GMT; Path=/
6) The browser requests "http://sunmail.domain.ru/uwc/auth" using the newly set cookie (looks like the old one to me though):
Cookie: amlbcookie=02; iPlanetDirectoryPro=AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#
7) The "sunmail" web-server checks the AM session validity with the same "psam.domain.ru". It sends a series of POSTs to /amserver/namingservice:
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<RequestSet vers="1.0" svcid="com.iplanet.am.naming" reqid="685">
<Request><![CDATA[
<NamingRequest vers="1.0" reqid="324" sessid="AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#">
<GetNamingProfile>
</GetNamingProfile>
</NamingRequest>]]>
</Request>
</RequestSet>(receives a large XML list of different Access Manager configuration parameters and URLs)
...then a double-request to /amserver/sessionservice:
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<RequestSet vers="1.0" svcid="Session" reqid="686">
<Request><![CDATA[
<SessionRequest vers="1.0" reqid="678">
<GetSession reset="true">
<SessionID>AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#</SessionID>
</GetSession>
</SessionRequest>]]>
</Request>
<Request><![CDATA[
<SessionRequest vers="1.0" reqid="679">
<AddSessionListener>
<URL>http://sunmail.domain.ru:80/UpdateAgentCacheServlet?shortcircuit=false</URL>
<SessionID>AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1+xTqH7C3I=@AAJTSQACMDI=#</SessionID>
</AddSessionListener>
</SessionRequest>]]>
</Request>
</RequestSet>As a result it receives an XML with a lot of user-specific information (the username, LDAP DN, preferred locale, auth module used, etc.)
!!!*** Now, the problem part ***!!!
8) And then "sunmail" POSTs a broken cookie to "psam" (note the space in mid-text, where the "plus" sign was previously). As we know, "+" is often used in URLs to "escape" the space character. Perhaps some URL cleanup routine backfired here.
I have double-checked, it is not the reverse proxy on "psam" breaking things. It is "sunmail" (UWC/CE or Policy Agent, don't know for certain) supplying the broken request. I looked over the large XML responses to the two previous requests, whenever they mention the session cookie value, the "plus" is there.
For the most detail I can provide, I'll even paste the whole HTTP packet:
POST /amserver/sessionservice HTTP/1.1
Proxy-agent: Sun-Java-System-Web-Server/7.0
Cookie: iPlanetDirectoryPro=AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1 xTqH7C3I=@AAJTSQACMDI=#;amlbcookie=null
Content-type: text/xml;charset=UTF-8
Content-length: 336
Cache-control: no-cache
Pragma: no-cache
User-agent: Java/1.5.0_09
Accept: text/html, image/gif, image/jpeg, *; q=.2, */*; q=.2
Host: cos-psam-01.domain.ru
Client-ip: 194.xxx.xxx.xxx
Via: 1.1 https-weblb.domain.ru
Connection: keep-alive
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<RequestSet vers="1.0" svcid="session" reqid="258">
<Request><![CDATA[<SessionRequest vers="1.0" reqid="254">
<GetSession reset="true">
<SessionID>AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1 xTqH7C3I=@AAJTSQACMDI=#</SessionID>
</GetSession>
</SessionRequest>]]></Request>
</RequestSet> The server's error response is apparent:
HTTP/1.1 200 OK
Server: Sun-Java-System-Web-Server/7.0
Date: Thu, 31 Jul 2008 05:49:50 GMT
Content-type: text/html
Transfer-encoding: chunked
19b
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<ResponseSet vers="1.0" svcid="session" reqid="258">
<Response><![CDATA[<SessionResponse vers="1.0" reqid="254">
<GetSession>
<Exception>AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1 xTqH7C3I=@AAJTSQACMDI=# Invalid session ID
AQIC5wM2LY4SfcwuT2ASCrsfO78nXdceEHXeH1 xTqH7C3I=@AAJTSQACMDI=#</Exception>
</GetSession>
</SessionResponse>]]></Response>
</ResponseSet>On the few occasions when the AM cookie contains no "plus" characters, the SSO works like a charm (also checked by a sniffer). Whenever there is a "plus", it breaks.
For reference, here's a working final request-response (one with a good cookie, as received by the load-balancer web-server). Request looks a bit different:
POST /amserver/sessionservice HTTP/1.1
Cookie: iPlanetDirectoryPro=AQIC5wM2LY4Sfcy/5sEzVmuq9z1ggdHOkBDgVFAwfhqvn4U=@AAJTSQACMDI=#;amlbcookie=null
Content-Type: text/xml;charset=UTF-8
Content-Length: 379
Cache-Control: no-cache
Pragma: no-cache
User-Agent: Java/1.5.0_09
Host: psam.domain.ru
Accept: text/html, image/gif, image/jpeg, *; q=.2, */*; q=.2
Connection: keep-alive
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<RequestSet vers="1.0" svcid="session" reqid="281">
<Request><![CDATA[<SessionRequest vers="1.0" reqid="277">
<SetProperty>
<SessionID>AQIC5wM2LY4Sfcy/5sEzVmuq9z1ggdHOkBDgVFAwfhqvn4U=@AAJTSQACMDI=#</SessionID>
<Property name="uwcstatus" value="active"></Property>
</SetProperty>
</SessionRequest>]]></Request>
</RequestSet> ...and the response is OK:
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<ResponseSet vers="1.0" svcid="session" reqid="281">
<Response><![CDATA[<SessionResponse vers="1.0" reqid="277">
<SetProperty>
<OK></OK>
</SetProperty>
</SessionResponse>]]></Response>
</ResponseSet>

There have been a few reports of the same behaviour with other customers - specifically with the handling of the encoding of "+" characters to " ". It relates to how cookie encoding/decoding is performed (as you have already observed).
The solution for these customers was the following:
=> AM server/client side:
Ensure that com.iplanet.am.cookie.encode=false in AMConfig.properties and AMAgent.properties on all systems.
=> AM client (UWC) side:
- Set <property name="encodeCookies" value="false"/> in /var/opt/SUNWuwc/WEB-INF/sun-web.xml. This will prevent UWC from trying to urldecode the cookie it receives and therefore stops it turning the + into a space e.g.
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE sun-web-app PUBLIC '-//Sun Microsystems, Inc.//DTD Sun ONE Application Server 7.0 Servlet 2.3//EN' 'file:///net/wajra.india.sun.com/export/share/dtd/sun-web-app_2_3-1.dtd'>
<sun-web-app>
   <property name="encodeCookies" value="false"/>
   <session-config>
      <session-manager/>
   </session-config>
   <jsp-config/>
<property name="allowLinking" value="true" />
</sun-web-app>Regards,
Shane.

TCAM and Access Lists

Similar Messages

Maybe you are looking for