FS#44008 - [linux] Starting userpace app with infinite recursion hangs kernel instead of SEGFAULT
Attached to Project:
Arch Linux
Opened by Lucjan B (lucck) - Monday, 02 March 2015, 08:55 GMT
Last edited by Doug Newgard (Scimmia) - Thursday, 13 August 2015, 02:50 GMT
Opened by Lucjan B (lucck) - Monday, 02 March 2015, 08:55 GMT
Last edited by Doug Newgard (Scimmia) - Thursday, 13 August 2015, 02:50 GMT
|
Details
Description:
Starting a program with infinite recursion (pls find attached source file test.c) on x86_64 kernel freeze the kernel. Symptoms: - Ctrl+C and killall -9 test doesn't work and application won't exit to the shell - After some time other tasks stopped responding - calling ps aux on another terminal freeze the ps command In the kernel dmesg sigfault is reported, but failed application is not interrrupted test[1614]: segfault at 7fffee656ffc ip 000000000040055e sp 00007fffee656ff0 error 6 in test[400000+1000] Additional info: * Linux 3.18.6.1 * Standard arch package kernel Steps to reproduce: - compile the application gcc test.c -o test - run application on the terminal ./test - try to intterrupt the application (CTRL+C) or killall -9 test (without any results) - try to run ps aux (It will cause ps hangs ) - after some time other tasks stopped responding randomly It affects only 64 bit kernel 32 bit kernel works properly (SEGFAULT interrupt the failed test program) 
test.c
(0.5 KiB)
|
This task depends upon
Closed by Doug Newgard (Scimmia)
Thursday, 13 August 2015, 02:50 GMT
Reason for closing: Fixed
Additional comments about closing: 3.19
Thursday, 13 August 2015, 02:50 GMT
Reason for closing: Fixed
Additional comments about closing: 3.19
I made some tests. Kernel 3.17 properly handle that case with SEGFAULT. Kernel 4.0rc also handle that case properly.
I think that this bug is related to:
http://seclists.org/oss-sec/2014/q4/777
Could you please provide provide a update package?
Thank you
Lucjan
Just a moment I'll recompile form source and I'll check.
I can confirm kernel 3.19 working properly. (It terminates task with SEGFAULT)
The reporter stated it was fixed in 3.19 and hasn't given any indication that it's still an issue.
> I am still getting this on 4.1.4-1.
Your machine locks when you compile / run the attached test case here?
> As a developer who runs into infinite loops every now and again this is really annoying.
It's not caused by infinite loops. It was caused by some unknown bug triggered by hitting the end of the main thread stack.
The attached source code does not freeze on this test case but on other infinite recursion (sorry I didn't mean loops) it does. The next time I find a test case I will open a new issue.