Cache-EX Loopback

[galiano]

I can not be sure that the cause is ecmhandler. I use the latest versions + cwc_patch, on four different devices in the home. Memory leak, occurs slowly, but it is fact for all platforms.

[CapNCooK]

Do you experience this memleaking also in oscams not-using cache-ex ?
If so, would you be able to produce a valgrind log showing the leak ?

(if you need help, let me know)

[CapNCooK]

Memleak fix in 8920... Can someone confirm this is now fixed ?

[Dimmie]

Memleak seems to be fixed but cache-ex loopback not (yet).
Also 8922 still takes much higher cpu load compared to 8894.

[CapNCooK]

@Dimmie:

We setup a full test-environment for this, but we _cannot_ replicate your issue...

I need to know more about your environment.. Which build you tested on, is it patched ?
Can you post the full configs from both oscams in test ?

--

About my environment;

- 1 Big CE, sending cache to TestOscam1
- TestOscam1 forwards cache to TestOscam2
- TestOscam2 returns cache to TestOscam1

On testoscam1, 0 cache is returned! (All tested with protocol: CCcam).

[Dimmie]

The config doesn't change, only after 8894 suddenly there is a cacheex loop and very high cpu usage..
All clients/readers are like this:


oscam.user

[account]
user = x-cacheex3
pwd = x
description = cache exchange mode 3
uniq = 3
keepalive = 0
allowedprotocols = cccam
group = 50
cacheex = 3
cacheex_maxhop = 2
cacheex_drop_csp = 1
cacheex_allow_request = 0
cccmaxhops = -1
cccstealth = 1
numusers = 0
penalty = 0


oscam.server

[reader]
label = x-cacheex3
description = cache exchange mode 3
protocol = cccam
device = x,x
user = x
password = x
inactivitytimeout = 30
cacheex = 3
cacheex_maxhop = 2
cacheex_allow_request = 0
cacheex_drop_csp = 1
group = 1,11,21,50,51,52,55,56,57,59,60,61
blockemm-unknown = 1
blockemm-u = 1
blockemm-s = 1
blockemm-g = 1
dropbadcws = 1
cccversion = 2.3.0
cccmaxhops = 0
ccckeepalive = 1
cccreconnect = 0
audisabled = 1


global settings:


[global]
logfile = /var/log/oscam.log
clienttimeout = 2000
fallbacktimeout = 1600
unlockparental = 1
nice = -1
maxlogsize = 1000
readerrestartseconds = 20
dropdups = 1
lb_nbest_readers = 2
lb_auto_betatunnel = 0
double_check_caid = 0100
failbantime = 60
failbancount = 3
double_check = 1


[cache]
delay = 15
max_time = 12
max_count = 7500
wait_time = 0:50:100,0100:130:180,0500:80:130,0602:1 20:170,0604:120:170,0919:10:25,0960:10:2 5,0961:10:25,0963:10:25,09C4:20:70,098C: 20:70,09CD:10:25,098C:20:70,1702:20:70,0 D00:80:130
csp_port = 64000
cwcycle_check_enable = 1
cwcycle_check_caid = 0100
cwcycle_maxlist = 2000
cwcycle_keeptime = 5
cwcycle_onbad = 1
cwcycle_dropold = 1
cwcycle_sensitive = 3


When you test you will also see that the amount of cache is suddenly extremely raised starting with build 8895.

I use VPS 5,8Ghz 1GB 20GB as external servers. Also tested local with 2 desktop pc's with debian 7 + oscam. Same problem.
To see the CPU usage rising just install oscam 8894 first, let it run and connect to the server with putty and the 'top' command. Then change oscam to a later build. You will notice the cpu load immediately. In oscam check cacheex size > it will be tripled when you compile above 8894.
Oscam is complied with this commandline: cmake -DCS_CACHEEX=1 -DWEBIF=1 IRDETO_GUESSING=1 -DCS_WITH_DOUBLECHECK=1 ..

[sat4ber]

Guys, I think that I have found the problem.

1) On 8894/8922 with no ECM/CW Cycle patch there is a cache reloop if you turn on the cycle check. When you use the ECM and CW Cycle patch (8894) there is no ccam loopback.
2) On 8894/8922 when you turn on the CCcam stealth mode you will create on your friends server a loopback and you receive the same cache back.

For example VPS1 (stealth on) VPS2 (stealth off) > cache loopback to VPS1.

I have tested with Oscam 8894 (ECM and CW cycle patch) 8894 and 8922 without the patch.

(If I change VPS1 to stealth mode you need to reboot VPS2).


Screenshot VPS1:

	Client	vps2	X	X	REVERSE CACHE PUSH	0	6	0	0	0
	Client	server1	X	X	REVERSE CACHE PUSH	0	6	0	0	0
	Reader	vps2	X	X	REVERSE CACHE PUSH	6	0	0	0	0

Only with the stealth mode you get the cache back.

[CapNCooK]

Dimmie,

You need to learn more about changes in 8895.
Max_Count has been removed in 8895!

Learn here:
=> http://www.streamboard.tv/wbb2/threa...605#post477605
=> http://www.streamboard.tv/wbb2/threa...392#post476392

So yes, size is higher, and yes.. with higher size CPU is higher.

--

Back to the loopback problem. I see you use multiple groups in readers.
Multiple groups with cache destroys de-duplication.

Can you make a simple test, like the test i did, having both user/reader in 1 group ?

Also, you did not answer all of my questions, which does not help.

[CapNCooK]

Guys, I think that I have found the problem.

Only with the stealth mode you get the cache back.

Yes! STEALTH cannot be used in CE, it needs extended!

[sat4ber]

@CapNCook

With Extended you mean in the user config?
Did you already now this problem? I think that not everyone is knowing this.