Created attachment 361214 [details] kernel panic Description of problem: kernel panic when boot rhel3.9-x86_64 on AMD host with smp or up kernel. Version-Release number of selected component (if applicable): kvm-83-105.el5_4.3 kvm-83-114.el5 How reproducible: always Steps to Reproduce: 1. Install rhel3.9-x86_64 on AMD host with iso or network or pxe. CLI: /usr/libexec/qemu-kvm -drive file=/data/images/rhel/rhel3.9-64-amd.raw,if=ide -dirve file=/data/images/rhel/boot.iso,index=2,media=cdrom -no-hpet -rtc-td-hack -smp 2 -m 4G -uuid `uuidgen` -net nic,model=e1000,macaddr=00:96:88:FE:DD:55,vlan=0 -net tap,vlan=0,script=/etc/qemu-ifup -cpu qemu64,+sse2 -boot d -monitor stdio -vnc :8 2. boot vm 3. Actual results: Expected results: attach the error Additional info: 1. this issue can be reproduced on kvm-83-105.el5_4.3 and kvm-83-114.el5 2. host: processor : 15 vendor_id : AuthenticAMD cpu family : 16 model : 2 model name : Quad-Core AMD Opteron(tm) Processor 8356 3. rhel3.9-x86_64 can be installed and booted successfully on Intel host.
Created attachment 361763 [details] Kernel panic: pci_map_single: high address but no IOMMU with 2gb ram. vm starts normally with 4gb ram. -> Kernel panic: pci_map_single: high address but no IOMMU. ( log is attached )
We don't emulate IOMMU and without it 4G of memory is not supported. Shouldn't work on Intel either. Can you provide dmesg on Intel host?
Actually vmware analyzed this problem for us: http://communities.vmware.com/thread/42332 "QA found that RedHat AS3 update2 and newer (up to and including latest available at this moment, update7) do not work with hugemem on AMD processors. It happens because RHAS3 amd64 kernels require either AMD's IOMMU, or they'll crash. Workaround for these kernels is (hopefully) to install *.ia32e.rpm package - like kernel-2.4.21-27.EL.ia32e.rpm. But they won't work out of the box."
Moving to on_qa. It should be closed as 'not a bug'
I check on StableSystem :x86-64-3as.ss.eng.bos.redhat.com, which memory is nearly 4G .qa.[root@x86-64-3as root]# free -m total used free shared buffers cached Mem: 3899 1344 2555 0 201 657 -/+ buffers/cache: 486 3413 Swap: 1992 0 1992 .qa.[root@x86-64-3as root]# uname -a Linux x86-64-3as.ss.eng.bos.redhat.com 2.4.21-63.ELsmp #1 SMP Wed Oct 28 23:07:38 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux .qa.[root@x86-64-3as root]# cat /proc/cpuinfo processor : 1 vendor_id : AuthenticAMD cpu family : 15 model : 5 model name : AMD Opteron(tm) Processor 246
What info do you need? Where is the question?
1. Guest can not boot with 4G mem 2. according to comment #5, rhel3.9 amd64 require either AMD's IOMMU or they'll crash when boot with mem more than 2G, right? 3. physical machine can boot with 4G mem. if 2 is correct, 3 is impossible. 3 is really exist,so 2 is incorrect.
(In reply to comment #7) > 1. Guest can not boot with 4G mem > 2. according to comment #5, rhel3.9 amd64 require either AMD's > IOMMU or they'll crash when boot with mem more than 2G, right? > 3. physical machine can boot with 4G mem. > > if 2 is correct, 3 is impossible. > 3 is really exist,so 2 is incorrect. Modify point 2 it should be: 4G not 2G 2. according to comment #5, rhel3.9 amd64 require either AMD's IOMMU or they'll crash when boot with 4G mem, right?
(In reply to comment #7) > 1. Guest can not boot with 4G mem > 2. according to comment #5, rhel3.9 amd64 require either AMD's > IOMMU or they'll crash when boot with mem more than 2G, right? > 3. physical machine can boot with 4G mem. > > if 2 is correct, 3 is impossible. > 3 is really exist,so 2 is incorrect. Physical machine obviously has AMD IOMMU. 2 is correct and 3 is correct.
*** Bug 654550 has been marked as a duplicate of this bug. ***