Hello Xenofon,
thanks for the additional information regarding the load_crl(..) function. Regarding the
question about making it configurable, I think the idea was to have a new configuration
parameter to disable the SSL context duplication per process, to make it easier testable
and provide a way to deactivate if its makes problems. It could be then maybe made the
default later on, if its works stable. It will be certainly some code duplication, but you
can probably comment better how easy or difficult it could be implemented.
Cheers,
Henning
--
Henning Westerholt –
https://skalatan.de/blog/
Kamailio services –
https://gilawa.com<https://gilawa.com/>
From: Xenofon Karamanos via sr-dev <sr-dev(a)lists.kamailio.org>
Sent: Dienstag, 14. Mai 2024 15:54
To: Kamailio (SER) - Development Mailing List <sr-dev(a)lists.kamailio.org>rg>;
miconda(a)gmail.com
Cc: Xenofon Karamanos <xk(a)gilawa.com>
Subject: [sr-dev] Re: TLS.reload and memory usage of SSL_CTX structs
Hello Daniel,
Thanks for the feedback!
Maybe i should have explained better regarding the leak. Indeed, i couldn't clearly
identify the leak itself but there is an indication that the leak is happening in the
SSL_CTX_load_verify_locations() function.
When i comment out that part of the load_crl() function there is no more exhaustion of
memory, so it's probably something SSL related.
Multiple issues suggest this also, look at:
https://groups.google.com/g/envoy-dev/c/JnWnH6HcsDU
https://github.com/pyca/pyopenssl/issues/1120
https://mta.openssl.org/pipermail/openssl-users/2015-April/001255.html
https://www.mail-archive.com/openssl-users@openssl.org/msg66240.html
https://www.mail-archive.com/openssl-users@openssl.org/msg57199.html
https://groups.google.com/g/mailing.openssl.users/c/R7bzJx167V4/m/lAGAPcDVm…
https://github.com/twisted/twisted/issues/12125
https://stackoverflow.com/questions/29845527/how-to-properly-uninitialize-o…
But i can't say for sure because the exact same function is called for CA list (maybe
exhaustion here is much slower).
I don't know the internal of SSL but when i enable the memory debugs, i see a lot of
mallocs and reallocs in that call, and no free() for the respective memories when
reloading each new tls.reload. Freeing() happens only when kamailio is terminated.
The idea of reusing instead of recreating came from this
https://stackoverflow.com/questions/67868098/how-to-duplicate-a-ssl-ctx-obj….
> (e.g., by libssl internally on handling traffic)
SSL_CTX should not be chagned once you have initialized an SSL connection according to
this
https://github.com/openssl/openssl/discussions/24203. maybe it was never needed but
then again i am not deeply familiar with how kamailio uses these ctx.
Make configurable what? whether to initialize it once and reuse it vs initialize per
process?
Thanks,
Xenofon
________________________________
From: Daniel-Constantin Mierla <miconda(a)gmail.com>
Sent: Tuesday, May 14, 2024 15:05
To: Kamailio (SER) - Development Mailing List <sr-dev(a)lists.kamailio.org>
Cc: Xenofon Karamanos <xk(a)gilawa.com>
Subject: Re: [sr-dev] TLS.reload and memory usage of SSL_CTX structs
Hello,
thanks for digging in it!
As I understand, this is somehow first about reducing the size of used memory, not clearly
identifying the leak itself.
The duplication done per process was to avoid races if local changes has to be performed
(e.g., by libssl internally on handling traffic), but maybe with the libssl3.x and new
threading approach is no longer needed, if ever was necessary.
What I would suggest is to make it configurable for now, so in case behaviour becomes
unstable on high load, one can switch between modes.
Cheers,
Daniel
On 14.05.24 10:36, Xenofon Karamanos via sr-dev wrote:
Hello all,
I am currently looking into issue
https://github.com/kamailio/kamailio/issues/3823
regarding the tls.reload and the constant increase in memory usage when it's called.
I tried to look something up related to CRL and related functions for what causing it but
nothing to obvious. The only culprit i could find was this function call in
SSL_verify_load_locations in
load_crl()<https://github.com/kamailio/kamailio/blob/9c2705b566103f4fad0…79>.
But the same exact function is also called in load_ca_list() with no such behaviour.
Next thing i notice, is that on the first rpc call of tls.reload, there is a
"significant" memory allocation.
What i mean by this is that when a fresh kamailio (5.8) is started up (with -m 7, for low
shared memory availability to try and replicate the issue and children=2) and then check
free shared memory usage using `kamcmd core.shmmem` is around 2196096 (bytes i am
assuming) for my system.
On the first call of tls.reload, it drops to 539248. That's a heavy drop so i started
digging why on so much allocation is caused. When loading a tls configuration,
ksr_fix_domain<https://github.com/kamailio/kamailio/blob/9c2705b566103f4…
allocates a new SSL_CTX per process. From what i am seeing this SSL_CTX, is basically the
same for each one, with same configuration without any difference on settings.
So, what i tried is to allocated it once (see tls_reload branch in my
repo<https://github.com/kamailio/kamailio/compare/master...xkaraman:kama…)d>),
and just share this context to all of the process and the memory used is way less and also
the unbounding memory seems to fixed or at least it's slowed that's barely
visible.
What are your thoughts on these? Is this something we can done? Is the assumption that tls
config will and be the same for every process or can one somehow differentiate from
another?
Thanks,
Xenofon
Some numbers for anyone interested:
Before the patch 5.8 branch and with settings such as -m 7 (shared memory) and children=2
(with children=8 it doesn't even start, failed to load cyphers in SSL_CTX):
Fresh kamailio start shared memory: total: 7340032
free: 2196096
After one tls.reload: free: 539248
After 100 tls.reload: free: 483776
After patch on master (kamailio -m 7 and children=2) (with children=8 it starts
normally):
Fresh kamailio start shared memory: total: 7340032
free: 3772000
After one tls.reload: free: 3690192
After 100 tls.reload: free: 3681056
[Image removed by
sender.]<https://github.com/kamailio/kamailio/compare/master...xkaraman:…
Comparing kamailio:master...xkaraman:tls_reload ·
kamailio/kamailio<https://github.com/kamailio/kamailio/compare/master...…
Kamailio - The Open Source SIP Server for large VoIP and real-time communication platforms
- - Comparing kamailio:master...xkaraman:tls_reload · kamailio/kamailio
github.com
[Image removed by
sender.]<https://github.com/xkaraman/kamailio/blob/9c2705b566103f4fad0ca…
kamailio/src/modules/tls/tls_domain.c at 9c2705b566103f4fad0cad63bc1e36ba3cca621f ·
xkaraman/kamailio<https://github.com/xkaraman/kamailio/blob/9c2705b56610…
Kamailio - The Open Source SIP Server for large VoIP and real-time communication platforms
- - xkaraman/kamailio
github.com
[Image removed by
sender.]<https://github.com/kamailio/kamailio/blob/9c2705b566103f4fad0ca…
kamailio/src/modules/tls/tls_domain.c at 9c2705b566103f4fad0cad63bc1e36ba3cca621f ·
kamailio/kamailio<https://github.com/kamailio/kamailio/blob/9c2705b56610…
Kamailio - The Open Source SIP Server for large VoIP and real-time communication platforms
- - kamailio/kamailio
github.com
[Image removed by
sender.]<https://github.com/kamailio/kamailio/issues/3823>
Memory usage increases everytime tls.reload is executed · Issue #3823 ·
kamailio/kamailio<https://github.com/kamailio/kamailio/issues/3823>
Description We are using Kamailio 5.7.4 on Debian 12 (from
http://deb.kamailio.org/kamailio57) with rtpengine as an Edgeproxy for our clients. The
instance terminates SIP/TLS (with Cliencertificate...
github.com
_______________________________________________
Kamailio (SER) - Development Mailing List
To unsubscribe send an email to
sr-dev-leave@lists.kamailio.org<mailto:sr-dev-leave@lists.kamailio.org>
--
Daniel-Constantin Mierla (@
asipto.com)
twitter.com/miconda --
linkedin.com/in/miconda
Kamailio Consultancy, Training and Development Services --
asipto.com