Re: [Cppcms-users] Segmentation fault on FreeBSD in high load

[Vinicius P. <ds...@ha...>]

Em 07/08/2012, às 01:55, augustin escreveu:

> On Tuesday, August 07, 2012 11:57:14 AM Artyom Beilis wrote:
>> 1  0x0000000800939cb3 in cppcms::http::response::set_content_header
>> (this=0x80ba7ac00, content_type=@0x7fffffffd590) at info.h:83
>> 
>> It looks like content_type you pass in info.h line 83 has very strange
>> address 0x7fffffffd590 - i.e. it is very close to 0x7FFFFFFFF It seems
>> strange to me check your code.
> 
> 
> High load + high memory address => running out of memory? Memory leak? Loop in 
> the code filling memory???
> 
> Check the memory usage on the server. See also what valgrind says...
> 

valgrind says there's some leaks and a lot of Mismatched free(), but I think it's on FreeBSD libs:

==74403== 12 bytes in 1 blocks are definitely lost in loss record 26 of 153
==74403==    at 0x102110B: malloc (in /usr/local/lib/valgrind/vgpreload_memcheck-amd64-freebsd.so)
==74403==    by 0x2BD7DFD: pthread_mutexattr_init (in /basejail/lib/libthr.so.3)
==74403==    by 0x172D42B: booster::recursive_mutex::recursive_mutex() (pthread.cpp:131)

==74403== Mismatched free() / delete / delete []
==74403==    at 0x1020EDE: free (in /usr/local/lib/valgrind/vgpreload_memcheck-amd64-freebsd.so)
==74403==    by 0x13B52E3: cppcms::crypto::key::reset() (crypto.cpp:423)
==74403==    by 0x139767C: cppcms::sessions::impl::hmac_factory::~hmac_factory() (hmac_encryptor.h:23)
==74403==    by 0x1390AF6: cppcms::session_pool::~session_pool() (memory:259)
==74403==    by 0x12D5010: cppcms::impl::service::~service() (memory:259)
==74403==    by 0x12D5440: cppcms::service::~service() (hold_ptr.h:27)

==74403== Mismatched free() / delete / delete []
==74403==    at 0x1020EDE: free (in /usr/local/lib/valgrind/vgpreload_memcheck-amd64-freebsd.so)
==74403==    by 0x1C5260C: std::locale::~locale() (in /basejail/usr/lib/libstdc++.so.6)
==74403==    by 0x12D4F7E: cppcms::impl::service::~service() (service.cpp:991)
==74403==    by 0x12D5440: cppcms::service::~service() (hold_ptr.h:27)

I never used valgrind since 2 days ago, by this log there's problems in libstdc++, libthr and something in cppcms too.

In Gentoo it don't show any leaks, only some warnings like:

==14171== Use of uninitialised value of size 8
==14171==    at 0x52926C6: booster::aio::io_service::set_timer_event(booster::ptime const&, booster::callback<void (booster::system::error_code const&)> const&) (io_service.cpp:207)
==14171==    by 0x5290A63: booster::aio::deadline_timer::async_wait(booster::callback<void (booster::system::error_code const&)> const&) (deadline_timer.cpp:95)
==14171==    by 0x4FBD25F: cppcms::impl::cgi::http_watchdog::check(booster::system::error_code const&) (http_api.cpp:728)
==14171==    by 0x4FC24AF: cppcms::impl::cgi::http_creator::http_creator(booster::aio::io_service&, cppcms::json::value const&, std::string const&, int) (http_api.cpp:98)
==14171==    by 0x4FBD52F: cppcms::impl::cgi::http_api_factory(cppcms::service&, std::string, int, int) (http_api.cpp:740)
==14171==    by 0x4EE7B7D: cppcms::service::setup_acceptor(cppcms::json::value const&, int, int) (service.cpp:765)
==14171==    by 0x4EE80ED: cppcms::service::start_acceptor(bool) (service.cpp:823)
==14171==    by 0x4EEB11C: cppcms::service::run() (service.cpp:557)
==14171==    by 0x406D99: main (wep.cpp:167)

Sorry for all this verbose logs in the list, if this doesn't help let me know please 

> Just some thoughts in passing...
> 
> Good luck,
> 
> Augustin.
> 

Thanks for your time,
Vinicius

> 
> 
> -- 
> Friends: http://www.reuniting.info/
> My projects:
> http://astralcity.org/ http://lesenjeux.fr/ http://linux.overshoot.tv/ 
> http://overshoot.tv/ http://charityware.info/ http://masquilier.org/
> http://openteacher.info/ http://minguo.info/ http://jacqueslemaire.fr/ 
> http://www.wechange.org/ http://searching911.info/
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> .
> 
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and 
> threat landscape has changed and how IT managers can respond. Discussions 
> will include endpoint security, mobile security and the latest in malware 
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users