On 05/29/2014 02:27 PM, Linus Torvalds wrote:
On Thu, May 29, 2014 at 2:07 PM, Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On Thu, May 29, 2014 at 11:21:09AM -0700, Greg Kroah-Hartman wrote:
On Thu, May 29, 2014 at 08:06:10AM -0700, Guenter Roeck wrote:
alpha:defconfig - Fails with 'Inconsistent kallsyms data'. New failure; need to debug.
Bisect points to commit 'vlan: Fix lockdep warning with stacked vlan devices' as culprit
for the alpha:defconfig build failure. Reverting this patch fixes the problem.
I looked at the patch, but I have no idea what may cause the failure.
Odd. Is this also broken in Linus's tree?
Agreed, it is odd. As I said, not idea what is causing this.
v3.15-rc7 is fine.
The bisection is almost certainly bogus, and this 'inconsistent
kallsyms data' issue is not new. I thought we fixed it, but it's
possible that there are alpha-specific issues and/or that it depends
on binutils versions.
In particular, when we've seen it in the past, it has depended on
exact alignment of code or data, which is why random changes
(particularly to percpu data) can seem to trigger it. Basically, two
symbols getting the same address and then being sorted differently by
'nm' or similar.
See for example
http://marc.info/?l=linux-next&m=133267456809502&w=2
where it hits ARM and rmk actually debugged that particular case.
Adding Michal to the cc, because I have a dim memory of this being
fixed, but I can't remember how. I hope Michal does.. Or maybe the
fact that alpha is hit by this just means that I mis-remember.
For now I "solved" the build problem by adding KALLSYMS_EXTRA_PASS=1"
to alpha builds.
Guenter
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html