Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Fixed
-
Fix Version/s: OpenVZ-legacy
-
Component/s: Containers::Kernel
-
Security Level: Public
-
Environment:Operating System: Other Linux
Platform: x86_64 (AMD64)
-
External issue URL:
-
External issue ID:1815
Description
It appears that the VDSO issue is either not fully fixed or is back. Inside a Fedora 13/x86_64 container on an Owl-current system, "ltrace top" ends in:
strlen(" PID USER PR NI VIRT R"...) = 69
__vsnprintf_chk(0x60eee0, 2112, 1, 2112, 0x40a68b) = 20
memset(0x31b12d0, '\000', 3480) = 0x31b12d0
gettimeofday(0x7fff2c0434b0, 0x7fff2c0434c0 <unfinished ...>
--- SIGSEGV (Segmentation fault) ---
+++ killed by SIGSEGV +++
$ uname -mrs
Linux 2.6.18-238.5.1.el5.028stab085.2.owl2 x86_64
$ cat /etc/redhat-release
Fedora release 13 (Goddard)
".owl2" makes no relevant changes other than the compile fix submitted to you by Vasiliy, which you already merged into 028stab087.1. You could want to try to reproduce the issue on other/nearby kernel versions; I guess you have several machines with Fedora 13 containers ready for such a check. We will proceed with further tests as well. I just wanted to communicate the bad news to you sooner rather than later, even though I am not sure if the issue is OpenVZ or maybe Openwall specific. It is possible that we never tested "top" on Fedora 13 when we thought that the VDSO issue was gone. Other commands, such as "ps" and "ls -l", work fine now (these used not to work under Fedora 13 when the VDSO issue first appeared in a test kernel).
strlen(" PID USER PR NI VIRT R"...) = 69
__vsnprintf_chk(0x60eee0, 2112, 1, 2112, 0x40a68b) = 20
memset(0x31b12d0, '\000', 3480) = 0x31b12d0
gettimeofday(0x7fff2c0434b0, 0x7fff2c0434c0 <unfinished ...>
--- SIGSEGV (Segmentation fault) ---
+++ killed by SIGSEGV +++
$ uname -mrs
Linux 2.6.18-238.5.1.el5.028stab085.2.owl2 x86_64
$ cat /etc/redhat-release
Fedora release 13 (Goddard)
".owl2" makes no relevant changes other than the compile fix submitted to you by Vasiliy, which you already merged into 028stab087.1. You could want to try to reproduce the issue on other/nearby kernel versions; I guess you have several machines with Fedora 13 containers ready for such a check. We will proceed with further tests as well. I just wanted to communicate the bad news to you sooner rather than later, even though I am not sure if the issue is OpenVZ or maybe Openwall specific. It is possible that we never tested "top" on Fedora 13 when we thought that the VDSO issue was gone. Other commands, such as "ps" and "ls -l", work fine now (these used not to work under Fedora 13 when the VDSO issue first appeared in a test kernel).
