Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
That fixed the problem - I enabled the following in the kernel .config: CONFIG_KEYS=y CONFIG_KEYS_DEBUG_PROC_KEYS=y CONFIG_SECURITY=y CONFIG_SECURITY_NETWORK=y CONFIG_SECURITY_CAPABILITIES=y and the crash no longer happens. May I suggest that this be considered a bug such that openafs should check at compile time that the current kernel has this capability? Thanks. -- Andrew J Perrin - [EMAIL PROTECTED] - http://perrin.socsci.unc.edu Associate Professor of Sociology; Book Review Editor, _Social Forces_ University of North Carolina - CB#3210, Chapel Hill, NC 27599-3210 USA On Wed, 1 Oct 2008, Russ Allbery wrote: Andrew J Perrin [EMAIL PROTECTED] writes: I'm assuming this is Enable Security Keys, correct? I'll try it soon. I think so, yes. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
Andrew J Perrin [EMAIL PROTECTED] writes: That fixed the problem - I enabled the following in the kernel .config: CONFIG_KEYS=y CONFIG_KEYS_DEBUG_PROC_KEYS=y CONFIG_SECURITY=y CONFIG_SECURITY_NETWORK=y CONFIG_SECURITY_CAPABILITIES=y and the crash no longer happens. May I suggest that this be considered a bug such that openafs should check at compile time that the current kernel has this capability? Yeah, it is. We're already discussing that. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
Package: openafs-modules-source Version: 1.4.7.dfsg1-5 Severity: important -- System Information: Debian Release: lenny/sid APT prefers testing APT policy: (500, 'testing') Architecture: i386 (i686) Kernel: Linux 2.6.26 (SMP w/4 CPU cores) Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1) Shell: /bin/sh linked to /bin/bash Versions of packages openafs-modules-source depends on: ii bison 1:2.3.dfsg-5 A parser generator that is compati ii debhelper 7.0.15 helper programs for debian/rules ii flex2.5.35-2 A fast lexical analyzer generator. ii kernel-package 11.001-0.1 A utility for building Linux kerne ii module-assistant0.10.11.0tool to make module package creati openafs-modules-source recommends no packages. -- no debconf information When I compile the openafs modules using make-kpkg modules-image under debian lenny, openafs dies with a General Protection Fault between 1 and 24 hours after boot. This does *not* happen on the same machines running 2.6.24 but does happen running .25 and .26. The machines are (1) a Lenovo ThinkCentre with a Pentium 4HT processor; and (2) an IBM eSeries 226 with dual Xeon 2.8Ghz processors. This is the GPF: Sep 29 10:41:00 che kernel: general protection fault: [#1] SMP Sep 29 10:41:00 che kernel: Modules linked in: radeon drm openafs(P) rfcomm l2cap bluetooth video backlight output fan ac acpi_cpufreq cpufreq_ondemand cpufreq_powersave cpufreq_conservative cpufreq_userspace ide_generic usbhid usb_storage ide_cd_mod libusual snd_hda_intel snd_pcm_oss snd_mixer_oss piix ehci_hcd uhci_hcd ide_pci_generic i2c_i801 ide_core thermal snd_pcm usbcore snd_timer snd_page_alloc snd_hwdep intel_agp agpgart snd processor button soundcore Sep 29 10:41:00 che kernel: Sep 29 10:41:00 che kernel: Pid: 3485, comm: afsd Tainted: P (2.6.25 #1) Sep 29 10:41:00 che kernel: EIP: 0060:[c0368c67] EFLAGS: 00010282 CPU: 1 Sep 29 10:41:00 che kernel: EIP is at _read_lock+0x0/0xc Sep 29 10:41:00 che kernel: EAX: fffe EBX: 0010 ECX: 0002 EDX: Sep 29 10:41:00 che kernel: ESI: fffe EDI: 48e0e8fc EBP: 48e0e87c ESP: f72d9f7c Sep 29 10:41:00 che kernel: DS: 007b ES: 007b FS: 00d8 GS: SS: 0068 Sep 29 10:41:00 che kernel: Process afsd (pid: 3485, ti=f72d8000 task=f7fef5c0 task.ti=f72d8000) Sep 29 10:41:00 che kernel: Stack: f94c15c5 0010 f72d9fb8 f94c7b3c 48e0e832 48e0e8fc f94b660d 0001 Sep 29 10:41:00 che kernel:002f 48e0da6b 48e0e8a0 48e0da9a 48e0e8d7 007079a0 48e0e87c Sep 29 10:41:00 che kernel:0001 f72d5da0 f9505a96 f7fef7a8 f950dc37 007b Sep 29 10:41:00 che kernel: Call Trace: Sep 29 10:41:00 che kernel: [f94c15c5] afs_osi_TraverseProcTable+0x12/0x5e [openafs] Sep 29 10:41:00 che kernel: [f94c7b3c] afs_GCPAGs+0x8f/0x161 [openafs] Sep 29 10:41:00 che kernel: [f94b660d] afs_Daemon+0x4ae/0x50c [openafs] Sep 29 10:41:00 che kernel: [f9505a96] afsd_thread+0x484/0x64e [openafs] Sep 29 10:41:00 che kernel: [f9505612] afsd_thread+0x0/0x64e [openafs] Sep 29 10:41:00 che kernel: [c01038ef] kernel_thread_helper+0x7/0x18 Sep 29 10:41:00 che kernel: === Sep 29 10:41:00 che kernel: Code: d0 c3 f0 ff 00 31 d2 89 d0 c3 89 c2 f0 81 28 00 00 00 01 0f 94 c0 b9 01 00 00 00 84 c0 75 09 f0 81 02 00 00 00 01 31 c9 89 c8 c3 f0 83 28 01 79 05 e8 3a fd ff ff c3 9c 5a fa b9 00 01 00 00 f0 Sep 29 10:41:00 che kernel: EIP: [c0368c67] _read_lock+0x0/0xc SS:ESP 0068:f72d9f7c Sep 29 10:41:00 che kernel: ---[ end trace 64f7b129a232e344 ]--- -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
Andrew Perrin [EMAIL PROTECTED] writes: Kernel: Linux 2.6.26 (SMP w/4 CPU cores) This would seem to indicate a hand-built kernel rather than the stock Debian kernel. Is that correct? When I compile the openafs modules using make-kpkg modules-image under debian lenny, openafs dies with a General Protection Fault between 1 and 24 hours after boot. This does *not* happen on the same machines running 2.6.24 but does happen running .25 and .26. The machines are (1) a Lenovo ThinkCentre with a Pentium 4HT processor; and (2) an IBM eSeries 226 with dual Xeon 2.8Ghz processors. This is the GPF: [...] Sep 29 10:41:00 che kernel: Call Trace: Sep 29 10:41:00 che kernel: [f94c15c5] afs_osi_TraverseProcTable+0x12/0x5e [openafs] This function is empty when AFS is built with keyring support, since it then uses the kernel keyring mechanisms to clean up the PAGs rather than needing to do this pass. That's the main reason why I asked above about the kernel, since the stock Debian kernels have keyrings enabled, which should mean that an OpenAFS client built against them could not have the above crash. If you're building your own kernels, be sure to enable keyrings, since AFS takes advantage of them to do much better PAG management. That will probably resolve your problem. (There's likely still a bug here, but it's a bug in code that's increasingly becoming dead as people switch to using keyring-based PAGs. That's what the Linux kernel developers would prefer that we use, since it involves fewer ugly hacks.) -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
On Wed, 1 Oct 2008, Russ Allbery wrote: Andrew Perrin [EMAIL PROTECTED] writes: Kernel: Linux 2.6.26 (SMP w/4 CPU cores) This would seem to indicate a hand-built kernel rather than the stock Debian kernel. Is that correct? Yes, that's correct, they're hand built. When I compile the openafs modules using make-kpkg modules-image under debian lenny, openafs dies with a General Protection Fault between 1 and 24 hours after boot. This does *not* happen on the same machines running 2.6.24 but does happen running .25 and .26. The machines are (1) a Lenovo ThinkCentre with a Pentium 4HT processor; and (2) an IBM eSeries 226 with dual Xeon 2.8Ghz processors. This is the GPF: [...] Sep 29 10:41:00 che kernel: Call Trace: Sep 29 10:41:00 che kernel: [f94c15c5] afs_osi_TraverseProcTable+0x12/0x5e [openafs] This function is empty when AFS is built with keyring support, since it then uses the kernel keyring mechanisms to clean up the PAGs rather than needing to do this pass. That's the main reason why I asked above about the kernel, since the stock Debian kernels have keyrings enabled, which should mean that an OpenAFS client built against them could not have the above crash. I'm assuming this is Enable Security Keys, correct? I'll try it soon. If you're building your own kernels, be sure to enable keyrings, since AFS takes advantage of them to do much better PAG management. That will probably resolve your problem. (There's likely still a bug here, but it's a bug in code that's increasingly becoming dead as people switch to using keyring-based PAGs. That's what the Linux kernel developers would prefer that we use, since it involves fewer ugly hacks.) -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- Andrew J Perrin - [EMAIL PROTECTED] - http://perrin.socsci.unc.edu Associate Professor of Sociology; Book Review Editor, _Social Forces_ University of North Carolina - CB#3210, Chapel Hill, NC 27599-3210 USA -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#500850: openafs-modules-source: General Protection Fault under 2.6.25 and 2.6.26
Andrew J Perrin [EMAIL PROTECTED] writes: I'm assuming this is Enable Security Keys, correct? I'll try it soon. I think so, yes. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]