Comment 16 for bug 1497428

Revision history for this message
Nelson Elhage (nelhage) wrote :

Hi @ddstreet, thanks for the update.

We unfortunately weren't able to reproduce this on your test kernel, and have since moved to a newer kernel version for other reasons.

However, I can confirm that on the affected machine types, and only the affected machine type, we see a memory range in `/proc/iomem` that ends off of a multiple of 2M. Again, we do not see this on any other machines.

I've attached `/proc/iomem` and `/proc/zoneinfo` from an affected machine (currently running an LTS backport kernel:
Linux [redacted] 3.19.0-33-generic #38~14.04.1-Ubuntu SMP Fri Nov 6 18:17:28 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
)

I'm pretty sure they show the kind of mismatch you're talking about; iomem shows

100000000-f0fffefff : System RAM
f0ffff000-f0fffffff : RAM buffer
f10000000-f17ffffff : System RAM

But the “Node 0 Normal” zone, judging from (start_pfn, start_pfn+spanned) spans 100000000-f18000000. The machine in question is a c4.8xlarge in EC2.