Red Hat Bugzilla – Bug 120015
(FS PROC)Oops in proc_delete_inode
Last modified: 2007-11-30 17:10:39 EST
Description of problem: For some time already I am getting this annoying oops in all kernels that were released for FC1. The timings are different, it might oops on the same day or in several weeks. Oops is not fatal and may repeat over time. Unable to handle kernel NULL pointer dereference at virtual address 0000008d printing eip: c0167b29 *pde = 00000000 Oops: 0002 parport_pc lp parport 8139too mii ip_conntrack_ftp ip_nat_ftp ipt_MASQUERADE iptable_nat ip_conntrack ipt_REJECT iptable_filter ip_tables microcode keybdev mo CPU: 0 EIP: 0060:[<c0167b29>] Not tainted EFLAGS: 00010206 EIP is at proc_delete_inode [kernel] 0x29 (2.4.22-1.2174.nptl) eax: 0000007d ebx: c094ac80 ecx: 00000020 edx: c2b0e000 esi: c0167b00 edi: 00000000 ebp: c2b0ff90 esp: c2b0ff44 ds: 0068 es: 0068 ss: 0068 Process cpuinfo.pl (pid: 458752, stackpage=c2b0f000) Stack: c015b3db c094ac80 c094ac80 00000000 c0489280 c094ac80 00000fff c01585a3 c094ac80 bfe985a0 c2b0ff90 c094ac80 c014f055 c0489280 0000000d c014bdde c2b0ff90 bfe985a0 00000fff c0489280 c10ff300 c39fdb00 00000000 c08b3180 Call Trace: [<c015b3db>] iput [kernel] 0x12b (0xc2b0ff44) [<c01585a3>] dput [kernel] 0xb3 (0xc2b0ff60) [<c014f055>] path_release [kernel] 0x15 (0xc2b0ff74) [<c014bdde>] sys_readlink [kernel] 0x6e (0xc2b0ff80) [<c0109747>] system_call [kernel] 0x33 (0xc2b0ffc0) Code: ff 48 10 8b 42 24 83 48 14 08 89 54 24 04 e9 54 ff ff ff 8d Version-Release number of selected component (if applicable): Linux tank.linuxhacker.ru 2.4.22-1.2174.nptl #1 Wed Feb 18 16:38:32 EST 2004 i686 i686 i386 GNU/Linux How reproducible: I just have mrtg activated and use it with scripts like linux_stats.pl and cpuinfo.pl from contribs, all the oopses happened in scripts, though once it happened in mrtg itself. All the oopses have same backtrace. Perhaps another point that might influence the problem is that I enabled large pid support (kernel.pid_max = 2147483647 in sysctl.conf) Also all pid's that caused oopses have a pattern in them, all of them have lower 16 bits as zero. Here are all pids that had oopses: 196608 262144 589824 786432 1245184 65536 458752 Hopefully this is enough of info.
> Perhaps another point that might influence the problem is that I > enabled large pid support (kernel.pid_max = 2147483647 in sysctl.conf) ehhh never do that. FC1 doesn't support more than 32k or 64k pids anyway... things go boom if you do that.
Well, I sort of expected there might be some problems, but since this functionality was included into FC1 kernel I decided to give it at least some testing. And this particular oops is not all that hard to fix I guess (probably wrong unsigned int -> unsigned short conversion somewhere?), I would fix it myself except downloading src rpm for this kernel would take ages on my shitty link. And if you do not want people to use this large pid support in FC1 kernels, you should not export /proc/sys/kernel/pid_max in first place.
Thanks for the bug report. However, Red Hat no longer maintains this version of the product. Please upgrade to the latest version and open a new bug if the problem persists. The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases, and if you believe this bug is interesting to them, please report the problem in the bug tracker at: http://bugzilla.fedora.us/