Hi Lukas, Thank you for your reply. This frequent crash is due to a PTE with a value of 0xF000 CXXX XXXX C4BE. Total eight dumps were examined and all of them show the same format of PTE. I think the latest kernel may well still has this bug in it. But I have no idea how this entry come to exist. Either the 0xF in the most significant nibble of a PTE is a bad one, or the "#define __swp_offset(x) ((x).val >> SWP_OFFSET_SHIFT)" should mask off the most significant bits from M to 63, where M is the maximum number of physical address lines. 0xF in the most significant nibble of a PTE is probably valid, because of the NX bit and the Protection Key features of Intel paging entries. Thanks. Dashi Cao -----Original Message----- From: linux-x86_64-owner@xxxxxxxxxxxxxxx [mailto:linux-x86_64-owner@xxxxxxxxxxxxxxx] On Behalf Of Odzioba, Lukasz Sent: Monday, November 14, 2016 7:20 PM To: Dashi DS1 Cao <caods1@xxxxxxxxxx> Cc: 'linux-x86_64@xxxxxxxxxxxxxxx' <linux-x86_64@xxxxxxxxxxxxxxx>; 'linux-numa@xxxxxxxxxxxxxxx' <linux-numa@xxxxxxxxxxxxxxx> Subject: RE: Kernel crashes in __migration_entry_wait On Sunday, November 13, 2016 1:40 PM Dashi Cao wrote: > A X86_64 server repeatedly dumps once a while with the following signature: > (snip) > > KERNEL: vmlinux > DUMPFILE: 127.0.0.1-2016-10-03-09:59:36/vmcore [PARTIAL DUMP] > CPUS: 32 > DATE: Mon Oct 3 10:13:22 2016 > UPTIME: 4 days, 17:04:52 > LOAD AVERAGE: 0.49, 0.26, 0.24 > TASKS: 657 > NODENAME: node04-priv > RELEASE: 3.10.0-327.el7.x86_64 > (snip) > > It seems that this is a bug. I'm not sure if it has been identified and removed, but it cannot be found on the web. The customer was adviced to disable numa balancing to work around and I'm waiting for the latest results from them. Hi Dashi, Thank you for your report ,but this seems to be kernel from RedHat 7.2 not our latest one nor stable, so I am not sure how many people here may be interested in your issue. If you don't get answer you can talk to RH support. Also this kernel is not the latest available for 7.2 so you may just try to update it. Thanks, Lukas -- To unsubscribe from this list: send the line "unsubscribe linux-x86_64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-x86_64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
