Public bug reported: Problem Description ================================== I attempted to inject a Core Unit Checkstop error by flipping Core FIR bit 5 on a K80 (Nvidia) 42L Server ( FSP - gp4fp1.aus.stglabs.ibm.com ). I was expecting the server to crash due to a Sapphire assert ( PEL with SRC BB821410 ) and a Sapphire dump to be collected in the process.
However on injecting the error , the host ( Ubuntu NV ) crashed, and never recovered. OPAL however seemed to stay up and there were a plethora of mail box errors - B182953C logged by the FSP. Error Inject -------------- $ putscom pu.ex 10013100 5 1 1 -ib -p3 -c5 s1.ex k0:n0:s0:p03:c5 ecmd_ppc putscom pu.ex 10013100 5 1 1 -ib -p3 -c5 Console message -------------- [htx@gp4p01] [1m/sys/devices/system/cpu# [0m [ 1652.976253] Fatal Hypervisor Maintenance interrupt [Not recovered] [ 1652.976332] Error detail: Malfunction Alert [ 1652.976402] HMER: 8040000000000000 [ 1652.976450] Kernel panic - not syncing: Unrecoverable HMI exception [ 1652.976467] CPU: 24 PID: 1261 Comm: kworker/24:1 Tainted: P OE 3.16.0-37-generic #51~14.04.1-Ubuntu [ 1652.976530] Workqueue: events hmi_event_handler [ 1652.976561] Call Trace: [ 1652.976571] [c0000000189bf9e0] [c000000000017330] show_stack+0x170/0x290 (unreliable) [ 1652.976647] [c0000000189bfac0] [c0000000009eb8e4] dump_stack+0x90/0xbc [ 1652.976674] [c0000000189bfaf0] [c0000000009e2b5c] panic+0x104/0x2a8 [ 1652.976703] [c0000000189bfb80] [c00000000007306c] hmi_event_handler+0x19c/0x2b0 [ 1652.976732] [c0000000189bfc50] [c0000000000d62dc] process_one_work+0x1ac/0x4d0 [ 1652.976772] [c0000000189bfce0] [c0000000000d6b80] worker_thread+0x190/0x630 [ 1652.976800] [c0000000189bfd80] [c0000000000e0024] kthread+0x114/0x140 [ 1652.976837] [c0000000189bfe30] [c00000000000a468] ret_from_kernel_thread+0x5c/0x74 [ 1652.977085] ---[ end Kernel panic - not syncing: Unrecoverable HMI exception . . . . |------------------------------------------------------------------------------| | 0x50227CFC 06/14/2015 22:27:43 System Hypervisor Firmware mbox | | 0x50227CFC Processed Predictive Error B182953C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227CDD 06/14/2015 22:27:21 System Hypervisor Firmware spif | | 0x50227CDD Processed Predictive Error B182951C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227CC0 06/14/2015 22:26:57 System Hypervisor Firmware mbox | | 0x50227CC0 Processed Predictive Error B182953C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C9E 06/14/2015 22:26:31 System Hypervisor Firmware spif | | 0x50227C9E Processed Predictive Error B182951C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C88 06/14/2015 22:26:12 System Hypervisor Firmware mbox | | 0x50227C88 Processed Predictive Error B182953C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C7E 06/14/2015 22:26:06 System Hypervisor Firmware spif | | 0x50227C7E Processed Predictive Error B182951C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C4F 06/14/2015 22:25:27 System Hypervisor Firmware mbox | | 0x50227C4F Processed Predictive Error B182953C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C40 06/14/2015 22:25:16 System Hypervisor Firmware spif | | 0x50227C40 Processed Predictive Error B182951C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C17 06/14/2015 22:24:42 System Hypervisor Firmware mbox | | 0x50227C17 Processed Predictive Error B182953C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x50227C01 06/14/2015 22:24:25 System Hypervisor Firmware spif | | 0x50227C01 Processed Predictive Error B182951C | --> Unexpected mail box error , needs investigation |------------------------------------------------------------------------------| | 0x501CF14F 06/11/2015 15:32:31 Processor Unit (CPU) prdf | | 0x501CF14F Processed Predictive Error B113E504 | --> Error log corresponding to injected Core Unit Checkstop error ( Core FIR [5] ) |------------------------------------------------------------------------------| == Comment: #3 - MAHESH J. SALGAONKAR <[email protected]> - 2015-06-16 14:06:56 == Ah! I see that panic timeout is set to 0 (zero). That means kernel will wait forever after panic. The behaviour reported in this BUG is as expected when panic timeout is set to 0. Hence it is not a BUG. =================== root@gp4p01:/usr/lib/debug/boot# cat /proc/sys/kernel/panic 0 root@gp4p01:/usr/lib/debug/boot# echo 10 > /proc/sys/kernel/panic root@gp4p01:/usr/lib/debug/boot# cat /proc/sys/kernel/panic 10 root@gp4p01:/usr/lib/debug/boot# After setting panic_timeout to 10 seconds I see that system rebooted on unrecoverable HMI: [salgaonkarm@mars linux-2.6]$ fsp_cmd -i gp4fp1.aus.stglabs.ibm.com Checking if system 'gp4fp1.aus.stglabs.ibm.com' is accesible (ping test) spawn telnet gp4fp1.aus.stglabs.ibm.com Trying 9.3.136.91... Connected to gp4fp1.aus.stglabs.ibm.com. Escape character is '^]'. Linux 2.6.32-279.14.1.69.fsp_fld8_1.ppcnf-fsp2 (gp4fp1) (12:51 on Friday, 08 October 2004) gp4fp1 login: dev Password: $ smgr mfgState runtime $ $ putscom pu.ex 10013100 5 1 1 -ib -p3 -c5 s1.ex k0:n0:s0:p03:c5 ecmd_ppc putscom pu.ex 10013100 5 1 1 -ib -p3 -c5 $ smgr mfgState runtime $ smgr mfgState ipling <= System rebooting.. =================== For system to reboot after panic, please set panic timeout to non-zero value and try injecting core checkstop again. You can do that in two ways: 1. Boot kernel with "panic=<secs>" kernel option (See below for valid values) OR 2. Once OS is booted, echo non-zero value to /proc/sys/kernel/panic $ echo 10 > /proc/sys/kernel/panic Please refer to Documentation/kernel-parameters.txt for valid panic values: -----------------Documentation/kernel-parameters.txt---------------------------- panic= [KNL] Kernel behaviour on panic: delay <timeout> timeout > 0: seconds before rebooting timeout = 0: wait forever timeout < 0: reboot immediately Format: <timeout> -----------------Documentation/kernel-parameters.txt---------------------------- == Comment: #8 - Stewart Smith <[email protected]> - 2015-09-17 20:03:51 == You should be able to provide panic_timeout=180 as a kernel argument as a workaround. However, this is *completely* an ubuntu bug. Perhaps we want to modify the panic() handler in linux though. == Comment: #9 - Luciano Chavez <[email protected]> - 2015-09-18 10:47:31 == I was speaking to Feroz and Donna this morning (and sent a note about the same thing last night) and to them the issue is not the panic but that it does not reboot after it hits it. As explained there are three methods to override the panic timeout (add kernel.panic= to sysctl,conf, echo value to /proc/sys/kernel/panic, or pass panic= to kernel command line). So, I hope Feroz can chime in but they want the hard coded CONFIG_PANIC_TIMEOUT changed because the rationale customers would likely not be aware to change it despite the previously mentioned ways to do it. So, if that is what they want, we will have to send this to Canonical for their take on this. Hi Canconical, Can you please comment your take on this issue ? Thank you. ** Affects: ubuntu Importance: Undecided Assignee: Taco Screen team (taco-screen-team) Status: New ** Tags: architecture-ppc64le bugnameltc-126343 severity-high targetmilestone-inin14043 ** Tags added: architecture-ppc64le bugnameltc-126343 severity-high targetmilestone-inin14043 ** Changed in: ubuntu Assignee: (unassigned) => Taco Screen team (taco-screen-team) -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1512593 Title: :Ubuntu NV: Panic timeout=0 means Ubuntu does not reboot and recover from HMI (e.g. Core Unit Checkstop) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+bug/1512593/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
