> The pte value above is very puzzling -- it says that the page is not
> present, in which case the "writable" bit means PROT_NONE, not writable.
> The pte value is tested immediately after dereferencing the page table
> to make sure it's present, so I don't see how we could reach this point
> without the page being present.
>
> Maybe this is a compiler bug after all ...
Don't think so, Ganesh reported this problem on a 2.7.2.3 built kernel,
and two others (including myself) reported a problem of this nature on a
1.0.3a built kernel. It has to be something odd somewhere in the code.
Alternatively (a disturbing thought just occured to me), what if the
scenario is that 2.7.2.3 provokes a different problem that causes the
above, and 1.0.3a provokes a problem not identical to the above that
causes the VM problems. So maybe we are seeing the same problem caused by
different code sequencing? The only way to find this out for sure is to
compile the parts of the VM code as .S code and compare the outputs from
2.7.2.3 and 1.0.3a with the C source code to see what the hell is going
on.
Also we need a sure fire method of provoking the VM error (I can't provoke
it at the moment, but I'll keep trying to overload the machine so we
can get more detailed output).
Cheers,
Alex.
--- /\_/\ Legalise cannabis now! ( o.o ) Smoke some cannabis today! > ^ < Peace, Love, Unity and Respect to all.Check out http://www.tahallah.demon.co.ukA
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu