From: robl <ro...@mo...> - 2008-04-28 15:22:50
|
Hi, I've recently upgraded my UML setup from a 2.6.8+skas3 host kernel to a 2.6.18+skas on debian etch. I've also upgraded to using the nagafix kernel32-2.6.24.3 kernel to run my guest instances in. I uses to be able to specify mem=3072M to the guest instance, but when I try this now I see : robl@mekon:~/umltest$ /usr/bin/kernel32-2.6.24.3 udb0=./root_fs mem=3072M Core dump limits : soft - 0 hard - NONE Checking that ptrace can change system call numbers...OK Checking syscall emulation patch for ptrace...OK Checking advanced syscall emulation patch for ptrace...OK Checking for tmpfs mount on /dev/shm...OK Checking PROT_EXEC mmap in /dev/shm/...OK Checking for the skas3 patch in the host: - /proc/mm...found - PTRACE_FAULTINFO...found - PTRACE_LDT...found UML running in SKAS3 mode CONFIG_HIGHMEM not enabled - physical memory shrunk to 3045064704 bytes Segmentation fault If I boot with something like mem=2800M it boots ok. I have a tmpfs partion defined on the host : tmpfs 4.0G 1.3G 2.8G 31% /dev/shm (which is empty when no guest is running). I have the following set in my host kernel : CONFIG_HIGHMEM4G=y CONFIG_PAGE_OFFSET=0xC0000000 CONFIG_HIGHMEM=y CONFIG_PROC_MM=y Anybody have any ideas what might be setting this limit, as I'd expect to be able to specify up to mem=4096M Thanks, Rob |
From: Jeff D. <jd...@ad...> - 2008-04-28 16:31:29
|
On Mon, Apr 28, 2008 at 04:21:41PM +0100, robl wrote: > I've recently upgraded my UML setup from a 2.6.8+skas3 host kernel to a > 2.6.18+skas on debian etch. I've also upgraded to using the nagafix > kernel32-2.6.24.3 kernel to run my guest instances in. > > I uses to be able to specify mem=3072M to the guest instance, but when I > try this now I see : When did it work before? I fixed a related bug recently - the patch is below. Note that UML isn't completely robust against its memory bumping into its shared libraries and stack. The patch below plus linking UML statically should give you close to 3G of memory. > Anybody have any ideas what might be setting this limit, as I'd expect > to be able to specify up to mem=4096M Nope, on 32-bit you only have 3G of address space. Jeff -- Work email - jdike at linux dot intel dot com |
From: robl <ro...@mo...> - 2008-04-28 17:16:40
|
Jeff Dike wrote: > On Mon, Apr 28, 2008 at 04:21:41PM +0100, robl wrote: >> I've recently upgraded my UML setup from a 2.6.8+skas3 host kernel to a >> 2.6.18+skas on debian etch. I've also upgraded to using the nagafix >> kernel32-2.6.24.3 kernel to run my guest instances in. >> >> I uses to be able to specify mem=3072M to the guest instance, but when I >> try this now I see : > > When did it work before? I *think* I was using the linux2.6.16.18-bs2 kernel as my guest kernel previously (I've just updated from Debian sarge to etch, so I've had to recompile the host kernel to include skas support again). I performed the upgrade last night. > > I fixed a related bug recently - the patch is below. Note that UML > isn't completely robust against its memory bumping into its shared > libraries and stack. The patch below plus linking UML statically > should give you close to 3G of memory. Thanks, I'll give this a go. > >> Anybody have any ideas what might be setting this limit, as I'd expect >> to be able to specify up to mem=4096M > > Nope, on 32-bit you only have 3G of address space. Sorry, you're right I was using mem=3096M previously (although this is more than 3G though - I must have been lucky as it never went down). > > Jeff > |