Description of problem: If we are running kexec on a 32-bit machine, and /etc/sysconfig/kdump:KDUMP_KERNELVER is empty, and the machine has memory mapped above 4GB (you can tell by looking at /proc/iomem), then the kdump init script should automatically look for a PAE kernel, and then fall back to a non-PAE kernel if that is not found for the kexec kernel. The reason for this is the without the PAE kernel, a machine with > 4G won't be able to dump all of memory.
we should be able to get all of physical memory that the running kernel can access from the kdump kernel already, without needing to special case PAE. Do you have a case in which thats not working? If so, please provide me a reproducer.
Ping Chris?
Neil, Oops. I never looked back at this one, sorry. I'll have to try again; it certainly may have just been user error. I'll be away for a week, though, so I won't get back to it until the second week in October, though. Thanks for the reminder. Chris Lalancette
well, let me know when you finish retesting.
chris, any update on this? Its been sitting for some time.
Sorry Neil. I just tried this again, and I couldn't get it to fail; it seems that the kdump script is choosing the correct kernel for all of the cases I can come up with. I'll close this as NOTABUG. Thanks for being patient. Chris Lalancette