Description of problem: Specifying mem=XXX on NUMA ia64 boxes can lead to system panic. Version-Release number of selected component (if applicable): kernel-2.6.9-34.EL How reproducible: 100% Steps to Reproduce: 1. Boot large memory system (128GB) with mem=32GB on the boot cmdline Actual results: Uncompressing Linux... done Loading initrd initrd-2.6.9-34.EL.img...done Linux version 2.6.9-34.EL (bhcompile.redhat.com) (gcc version 3.4.5 20051201 (Red Hat 3.4.5-2)) #1 SMP Fri Feb 24 16:49:08 EST 2006 EFI v1.10 by INTEL: SALsystab=0x3002814b60 ACPI 2.0=0x300281b8e0 booting generic kernel on platform sn2 Number of logical nodes in system = 32 Number of memory chunks in system = 32 efi.trim_top: ignoring 8KB of memory at 0x2b400000000 due to granule hole at 0x2b452000000 Kernel panic - not syncing: pernode space for node 31 could not be allocated! Expected results: Boot should succeed. Additional info: Upstream patch fixes this: http://marc.theaimsgroup.com/?l=linux-ia64&m=109718964113183&w=2
While with a well-functioning system there's no reason for a customer to pass "mem=" to a machine, we do utilize that flag for working around and debugging other issues so really should ensure that it works on all architectures.
This has been broken all along in RHEL4 and is not a regression. Since no one has complained about it yet, I suggest pushing this into RHEL4 U5. P.