On Fri, Jun 12, 2020 at 04:47:57PM +1200, Michael Cree wrote:

On Thu, Jun 11, 2020 at 09:23:52PM -0700, Matt Turner wrote:

quoted

Since I noticed earlier that using maxcpus=1 on a 2-CPU system
prevented the system from hanging, I tried disabling CONFIG_SMP on my
1-CPU system as well. In doing so, I discovered that the RCU torture
module (RCU_TORTURE_TEST) triggers some null pointer dereferences on
Alpha when CONFIG_SMP is set, but works successfully when CONFIG_SMP
is unset.

That seems likely to be a symptom of the same underlying problem that
started this thread, don't you think? If so, I'll focus my attention
on that.

I wonder if that is related to user space segfaults we are now seeing
on SMP systems but not UP systems while building Alpha debian-ports.
It's happening in the test-suites of builds of certain software
(such as autogen and guile) but they always build successfully with
the test suite passing on a UP system.

When investigating I seem to recall it was a NULL (or near NULL)
pointer dereference but couldn't make any sense of how it might
have got into such an obviously wrong state.

By some miracle, I have avoided any experience with RCU bugs. ;) If
the RCU_TORTURE_TEST Oopses or the segfaults are repeatable and don't
go away with the WARN patch reverted, then perhaps it might be used to
bisect to something closer to the root cause?

Given the similarity to the SMP vs UP stuff and the RCU tests, I'd agree
that does seem like the best path to investigate.

-- 
Kees Cook