From: Dominique DELANDE (Dominique.Delande_at_spectro.jussieu.fr)
Date: Tue Jan 06 2009 - 06:43:36 PST
Paul, Sorry for the very late reply, I have been out of town in dec 2008. I applied the patch, which solves the problem: cr_restart now works fine. I tested with MVAPICH2-1.2p1. I will soon test OpenMPI too. Thanks a lot for your work: blcr is great software. Dominique Paul H. Hargrove wrote: > I have created a bugzilla entry for this issue at > http://upc-bugs.lbl.gov/bugzilla/show_bug.cgi?id=2454 > > I believe I have fixed this bug for the upcoming 0.8.0 beta release, but > can't be 100% certain my test case fully duplicates the circumstances of > the reported failure. So, I have also attached to the bugzilla entry a > patch relative to the 0.7.3 sources. I'd appreciate it if Dominique (or > anyone else who has seen this problem) could report success of failure > when using the patch. > > -Paul > > Paul H. Hargrove wrote: >> Dominique, >> I am now testing for a Beta release of 0.8.0 and I have been able to >> reproduce your Oops with the kernel version you indicated, and also >> with a vanilla 2.6.26 from kernel.org. The bug relates to the restore >> of the FPU state, and since none of our test suite perform any >> floating point math, the bug went unnoticed in our testing of 0.7.3 >> against the 2.6.26 kernel. >> >> I have a possible fix in testing right now, and it is one of the >> things on my list to finish before the 0.8.0 beta release. >> >> -Paul >> >> Paul H. Hargrove wrote: >>> Dominique, >>> >>> I must say that we have not seen an error like this before. >>> However, we've not done testing ourselves on a kernel more recent >>> than a vanilla 2.6.26. It seems likely that BLCR is not prepared for >>> something in the 2.6.26.6-49.fc8 kernel. >>> I hope to begin beta testing on BLCR 0.8.0 in late November, and we >>> are busy with other tasks between now and then. So, I am afraid >>> there is little chance that your problem will be resolved prior to that. >>> However, I will try to see that we include a kernel like your inour >>> testing of 0.8.0 to ensure we identify the cause of the problem and >>> get it solved as quickly as we can. >>> >>> -Paul >>> >>> >>> Dominique DELANDE wrote: >>>> Dear Sir or Madam, >>>> >>>> I am using blcr-0.7.3 to checkpoint MPI jobs, compiled either >>>> with MVAPICH2-1.2rc2 or OpenMPI-1.3b1. In both cases, >>>> checkpointing is apparently working, but restart ends >>>> with a kernel oops on each node running a MPI process (see below). >>>> The kernel oops is similar when using OpenMPI or MVAPICH2, I thus >>>> suspect that the problem comes from blcr. >>>> I am running Linux Fedora 8 with a 2.6.26 Fedora x86_64 kernel. >>>> Each node has two Dual Core AMD Opteron Processors 285 with 8 GBytes >>>> of memory, Infiniband DDR, and the OFED-1.4-rc3 stack. >>>> >>>> Any help will be appreciated. >>>> >>>> Best regards >>>> >>>> Dominique Delande >>>> >>>> ******************************* >>>> >>>> Ouput of 'uname -a': >>>> Linux node11.cluster.local 2.6.26.6-49.fc8 #1 SMP Fri Oct 17 >>>> 15:33:32 EDT 2008 x86_64 x86_64 x86_64 GNU/Linux >>>> >>>> Example: >>>> >>>> $ cr_restart context.20722 >>>> node11 kernel: Oops: 0002 [1] SMP >>>> node11 kernel: Code: 48 8b 11 31 c0 c3 48 83 e9 07 eb 00 31 d2 48 >>>> c7 c0 f2 ff ff ff c3 90 90 90 90 90 90 90 90 90 90 48 89 f8 89 d1 c1 >>>> e9 03 83 e2 07 <f3> 48 a5 89 d1 f3 a4 c3 66 66 66 66 2e 0f 1f 84 00 >>>> 00 00 00 00 >>>> node11 kernel: CR2: 0000000000000000 >>>> >>>> dmesg output is: >>>> >>>> BUG: unable to handle kernel NULL pointer dereference at >>>> 0000000000000000 >>>> IP: [<ffffffff81140d1b>] memcpy_c+0xb/0x20 >>>> PGD 22e4ed067 PUD 22dccf067 PMD 0 >>>> Oops: 0002 [1] SMP >>>> CPU 1 >>>> Modules linked in: blcr(U) blcr_vmadump(U) blcr_imports(U) nfs lockd >>>> nfs_acl sunrpc rdma_ucm(U) ib_sdp(U) rdma_cm(U) iw_cm(U) ib_addr(U) >>>> ib_ipoib(U) ib_cm(U) ib_sa(U) ipv6 ib_uverbs(U) ib_umad(U) >>>> mlx4_ib(U) mlx4_core(U) dm_mirror dm_log dm_multipath dm_mod >>>> cfi_cmdset_0002 cfi_util tg3 jedec_probe cfi_probe ib_mthca(U) >>>> gen_probe ib_mad(U) ck804xrom ib_core(U) mtd k8temp i2c_nforce2 >>>> i2c_core pcspkr chipreg hwmon shpchp map_funcs sr_mod cdrom sg >>>> pata_amd ata_generic pata_acpi sata_nv libata sd_mod scsi_mod ext3 >>>> jbd mbcache uhci_hcd ohci_hcd ehci_hcd [last unloaded: scsi_wait_scan] >>>> Pid: 20789, comm: cr_restart Not tainted 2.6.26.6-49.fc8 #1 >>>> RIP: 0010:[<ffffffff81140d1b>] [<ffffffff81140d1b>] memcpy_c+0xb/0x20 >>>> RSP: 0018:ffff81022e4c7770 EFLAGS: 00010246 >>>> RAX: 0000000000000000 RBX: 0000000000000200 RCX: 0000000000000040 >>>> RDX: 0000000000000000 RSI: ffff81022e4c7788 RDI: 0000000000000000 >>>> RBP: ffff81022e4c7b58 R08: 0000000000000000 R09: 0000000000000001 >>>> R10: 0000000000000000 R11: 0000000000000001 R12: ffff81022e4c7788 >>>> R13: ffff810128c2a240 R14: 0000000000000000 R15: ffffffffffffffff >>>> FS: 00007f4cabcdc6f0(0000) GS:ffff81022fa25300(0000) >>>> knlGS:0000000055788b90 >>>> CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 >>>> CR2: 0000000000000000 CR3: 000000012e96b000 CR4: 00000000000006e0 >>>> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 >>>> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 >>>> Process cr_restart (pid: 20789, threadinfo ffff81022e4c6000, task >>>> ffff81022e4596a0) >>>> Stack: ffffffffa0334524 0000000000000000 0000000000000000 >>>> 000000000000037f >>>> 0000000000000000 0000000000000000 0000ffff00009fc0 0000000000000000 >>>> 0000000000000000 0000000000000000 0000000000000000 0000000000000000 >>>> Call Trace: >>>> [<ffffffffa0334524>] ? :blcr_vmadump:vmadump_restore_cpu+0x197/0x5b4 >>>> [<ffffffff810aa7a0>] ? do_sync_read+0xe2/0x126 >>>> [<ffffffff810492dc>] ? autoremove_wake_function+0x0/0x38 >>>> [<ffffffff8100c12a>] ? system_call_after_swapgs+0x8a/0x8f >>>> [<ffffffffa0332048>] ? :blcr_vmadump:read_user+0x48/0x72 >>>> [<ffffffffa0333846>] :blcr_vmadump:vmadump_thaw_proc+0x169/0xb6b >>>> [<ffffffff8102fb8d>] ? hrtick_set+0xe0/0xe9 >>>> [<ffffffff8129748d>] ? thread_return+0x7e/0xab >>>> [<ffffffffa03455dc>] :blcr:cr_thaw_threads+0x18d/0x1f1 >>>> [<ffffffff810492dc>] ? autoremove_wake_function+0x0/0x38 >>>> [<ffffffffa0343413>] :blcr:cr_rstrt_child+0xa1e/0x1ca2 >>>> [<ffffffffa033e0b8>] ? :blcr:ctrl_ioctl+0x0/0x1bc >>>> [<ffffffffa033e22f>] :blcr:ctrl_ioctl+0x177/0x1bc >>>> [<ffffffff810e8aae>] proc_reg_unlocked_ioctl+0xa6/0xc6 >>>> [<ffffffff810b6aaa>] vfs_ioctl+0x2a/0x77 >>>> [<ffffffff810b6d45>] do_vfs_ioctl+0x24e/0x26b >>>> [<ffffffff8100a92a>] ? __switch_to+0xd8/0x376 >>>> [<ffffffff810b6db9>] sys_ioctl+0x57/0x7a >>>> [<ffffffff8100c12a>] system_call_after_swapgs+0x8a/0x8f >>>> >>>> >>>> Code: 48 8b 11 31 c0 c3 48 83 e9 07 eb 00 31 d2 48 c7 c0 f2 ff ff ff >>>> c3 90 90 90 90 90 90 90 90 90 90 48 89 f8 89 d1 c1 e9 03 83 e2 07 >>>> <f3> 48 a5 89 d1 f3 a4 c3 66 66 66 66 2e 0f 1f 84 00 00 00 00 00 >>>> RIP [<ffffffff81140d1b>] memcpy_c+0xb/0x20 >>>> RSP <ffff81022e4c7770> >>>> CR2: 0000000000000000 >>>> ---[ end trace 6c22587b3026d5be ]--- >>>> blcr: rstrt_watchdog: 'cr_restart' (tgid/pid 20788/20789) exited >>>> with signal 9 during restart >>>> >>>> >>> >>> >> >> > > -- Dominique Delande ([email protected]) Laboratoire Kastler-Brossel - Case 74 - Universite P. et M. Curie 4, place Jussieu, F-75252 Paris Cedex 05, FRANCE Phone : 33 (0)1 44 27 27 97 - Fax : 33 (0)1 44 27 38 45 Acces : Pyramide de la Scolarite Paris VI - 1er etage - Bureau 214