Sorry, you seem to have run in to a known but so far unfixed problem
that we are actively working on. We have been calling it the bigmem
problem since so far it seems to happen on 32+ gig memory machines. See
CR # 6654713. The symptom in all the related panics is a trashed kernel
page_t structure which leads to various panics depending on what the
page struct damage is.
Stu
Ian Blenke wrote:> I’m running SXCE b87 on a couple of Dells in an xVM cluster, and I just
> experienced this dom0 panic:
>
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 556024 kern.notice]
> page_get_pagecnt: out of range 6
> Jun 5 15:52:33 lab-xvm-00-a unix: [ID 100000 kern.notice]
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 655072 kern.notice]
> ffffff003c4e6d80 unix:page_get_pagecnt+38 ()
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 655072 kern.notice]
> ffffff003c4e6de0 unix:page_next_scan_large+4e ()
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 655072 kern.notice]
> ffffff003c4e6e60 genunix:fsflush_do_pages+c2 ()
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 655072 kern.notice]
> ffffff003c4e6ef0 genunix:fsflush+3b3 ()
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 655072 kern.notice]
> ffffff003c4e6f00 unix:thread_start+8 ()
> Jun 5 15:52:33 lab-xvm-00-a unix: [ID 100000 kern.notice]
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 672855 kern.notice] syncing
> file systems...
> Jun 5 15:52:33 lab-xvm-00-a unix: [ID 836849 kern.notice]
> Jun 5 15:52:33 lab-xvm-00-a ^Mpanic[cpu3]/thread=ffffff08e8f4d460:
> Jun 5 15:52:33 lab-xvm-00-a genunix: [ID 335743 kern.notice] BAD TRAP:
> type=e (#pf Page fault) rp=fffffffffbc960e0 addr=9 occurred in module
> "unix" due to a NULL pointer dereference
>
> This is a Dell 2970 dual Opteron 2220 dual core system with 32G of RAM.
> The dom0 network interfaces include onboard bnx and an add-in e1000
> (Intel 1000-PT).
>
> DomUs that were running at the time:
>
> * about a dozen 64bit paravirt Ubuntu Hardy (8.04) domU guests
> running with the XenSource 2.4.18 vmlinuz kernel from the Xen 3.1
> x86_64 binary tarball
> * one SXCE b85 paravirt domU
>
>
> The guests were all idle. Primarily playground/development guests with
> nobody using them at the time.
>
> Of the 32G of memory, only 12G was in use by those guests at the time,
> the dom0 still had 19G of memory unallocated (balloon).
>
> The domUs are all running on iSCSI devices exposed by the dom0’s
> initiator, as exposed from another Dell 2970 acting as an iSCSI server
> (zfs zvols, iscsishare=on).
>
> This is all in a lab, of course, none of this is production.
>
> Perhaps a b90 upgrade is in order. I’m willing to give just about
> anything a try.
>
> - Ian C. Blenke <ian@blenke.com> <iblenke@csdvrs.com>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> xen-discuss mailing list
> xen-discuss@opensolaris.org