[Libwebsockets] Valgrind error

Andy Green andy at warmcat.com
Thu Apr 26 02:40:49 CEST 2018



On 04/26/2018 12:27 AM, Brice Hamon wrote:
> Hi guys,
> 
> I am seen this when I valgrind my code.
> 
> ==2490== Conditional jump or move depends on uninitialised value(s)
> ==2490==    at 0x5A23AA6: ??? (in /lib64/libcrypto.so.1.0.0)
> ==2490==    by 0x5D47FEF: d2i_SSL_SESSION (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D2AEA0: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D2D081: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D410BA: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D16EE1: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D1B8CC: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x5D28DF3: ??? (in /lib64/libssl.so.1.0.0)
> ==2490==    by 0x56F3A59: lws_server_socket_service_ssl (in 
> /export/home/development/3rdparty/lib/libwebsockets.so.12)
> ==2490==    by 0x56F7B90: lws_adopt_descriptor_vhost (in 
> /export/home/development/3rdparty/lib/libwebsockets.so.12)
> ==2490==    by 0x56F8093: lws_server_socket_service (in 
> /export/home/development/3rdparty/lib/libwebsockets.so.12)
> ==2490==    by 0x56E8846: lws_service_fd_tsi (in 
> /export/home/development/3rdparty/lib/libwebsockets.so.12)
> ==2490==  Uninitialised value was created by a stack allocation
> ==2490==    at 0x59979F7: ??? (in /lib64/libcrypto.so.1.0.0)
> 
> It is happening in run time, and usually once, for a bunch of time then 
> nothing.
> 
> I don't know if it is a problem but I wanted to report it.
> 
> I am on Linux Opensuse
> Linux ydotm2 4.1.38-50-default #1 SMP PREEMPT Sun Feb 19 14:35:48 UTC 
> 2017 (6b4d8cb) x86_64 x86_64 x86_64 GNU/Linux
> with libwebsocket v2.4.0.

It seems to be happening during connection accept... unfortunately it's 
difficult with OpenSSL + valgrind to tell if that particular kind of 
report is us or OpenSSL + valgrind.

OpenSSL is known to read uninitialized stack deliberately

https://rachelbythebay.com/w/2012/12/14/quiet/

It looks like the one bt clue leads to this which openssl-dev say is a 
"false positive"

https://rt.openssl.org/Ticket/Display.html?id=3592&user=guest&pass=guest

So I think this is purely to do with OpenSSL and nothing to do with us 
in this case.

Although don't let it put you off telling us about any other valgrind 
reports, often they are the only way to understand there is a big 
problem hiding somewhere.

-Andy

> 
> Thank you,
> Brice.
> 
> 
> _______________________________________________
> Libwebsockets mailing list
> Libwebsockets at ml.libwebsockets.org
> https://libwebsockets.org/mailman/listinfo/libwebsockets
> 



More information about the Libwebsockets mailing list