Skip to contentSkip to conversation

Unable to handle kernel paging request in 2.6.16-rc1-mm4

DISCUSSION Lee Schermerhorn 1 participant 1 message Feb 1, 2006

lore
Share & tools
lore
#0 lore raw

Just a "heads up" because I haven't found reference to this in the ia64 nor kernel lists. Maybe just happens on my platform?

I'm seeing this on an HP rx8620 [16xIA64, 4 NUMA nodes, 32GB], booting subject kernel. Log excerpts:
---

Unable to handle kernel paging request at virtual address
a004037b4dc28110
events/12[62]: Oops 11012296146944 [1]
<snip>
Call Trace:
<snip>
 [<a00000010000cca0>] ia64_leave_kernel+0x0/0x280
                                spà000721fe84fb80 bspà000721fe849290
 [<a0000001001483b0>] free_block+0x90/0x360
                                spà000721fe84fd50 bspà000721fe849200
 [<a0000001001489d0>] drain_array_locked+0x110/0x180
                                spà000721fe84fd50 bspà000721fe8491b8
 [<a00000010014e6c0>] cache_reap+0x1a0/0x600
                                spà000721fe84fd50 bspà000721fe849140
 [<a0000001000c61c0>] run_workqueue+0x180/0x2c0
                                spà000721fe84fd60 bspà000721fe8490f0
 [<a0000001000c6540>] worker_thread+0x240/0x2c0
                                spà000721fe84fd60 bspà000721fe8490b8
<snip>
 BUG: events/12/62, lock held at task exit time!
 [a000000100867fe0] {cache_chain_mutex}

.. held by: events/12: 62 [e0000721fe848000, 110]
... acquired at: cache_reap+0x30/0x600
---
Eventually the boot hangs...

I did a bisection search of the mm4 broken out patches and isolated the problem to the "export-cpu-topology-by-sysfs" patches. I.e., if I comment these out, builds and boots fine [with a one-liner build fix from Christoph Lameter for an SGI sn pci source file]. Something overstepping a slab allocation? Still investigating.

Regards,
Lee