Hi, On 6/8/20 4:28 PM, Marc Zyngier wrote: > Recent changes made to Linux 5.8 have outlined that kvmtool > hardcodes the text offset instead of reading it from the arm64 > image itself. > > To address this, import the image header structure into kvmtool > and do the right thing. 32bit guests are still loaded to their > usual locations. > > While we're at it, check the image magic and default to the text > offset to be 0x80000 when image_size is 0, as described in the > kernel's booting.rst document. > > Cc: Alexandru Elisei <alexandru.elisei@xxxxxxx> > Reported-by: Ard Biesheuvel <ardb@xxxxxxxxxx> > Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx> > --- > > Notes: > v2: Check magic, assume offset to be 0x8000 if image_size is 0. > > [..] I wanted to review the patch, but then I noticed that Will merged it. So I settled for testing it, to make sure nothing breaks. I was able to reproduce the issue reported on the mailing list - without this patch, when the kernel is compiled with CONFIG_RANDOMIZE_BASE not set, the kernel doesn't boot; with this patch applied, the same kernel boots successfully. I also tested it with an aarch32 kernel, this is what I got: $ taskset -c 4,5 ./lkvm run -c2 -m1024 -k ../kvmtool-vms/zImage -d ../kvmtool-vms/debian-10.3.0.img -p 'earlycon root=/dev/vda2' # lkvm run -k ../kvmtool-vms/zImage -m 1024 -c 2 --name guest-1834 Warning: Kernel image magic not matching Warning: unable to translate host address 0x13115a4c82d05a4d to guest Fatal: kernel image too big to contain in guest memory. IMO, works as expected. Thanks, Alex _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
