Hi Bhupesh, -----Original Message----- > >> Yes, the ARMv8 simulator can support both ARMv8.2 LPA and LVA features > >> and I tested the above suggestion earlier also on the same and landed > >> into incorrect paddress calculation issues. > >> > >> Since the simulator is slow and its time taking to insert rebuilt > >> user-space components in the Linaro initramfs being used on the same, I > >> would suggest that we go ahead with the above code for now and later > >> when I have more results from the simulator/real hardware, I will send > >> out a follow-up patch (if needed) to fix the paddress calculation. > > > > Hmm, it theoretically needs __pte_to_phys(), right? > > So if you have an issue with it, there may be a bug somewhere > > and need to debug it. Do you have the detailed information? > > Its not very easy to get the detailed UART console logs from the > simulator, so it is hard to get all the debug logs from makedumpfile, so > I am trying debugging the issue via 'gdb' by adding it to the initramfs.
I don't know what environment the simulater is, but you cannot capture a vmcore with cp or something? then debug makedumpfile with it on a real arm64 machine. I usually do like this. # cp --sparse=always /proc/vmcore vmcore I'm willing to debug it if you would send me the vmcore and vmlinux. > > However till then to fix the regression reported with upstream > makedumpfile on arm64 platforms which don't support ARMv8.2-LPA > extensions (e.g. Cortex-A57) and run a newer kernel with PA=52-bit > configuration, we can apply this patch for now. > > I have tested this on non-ARMv8.2-LPA platforms like apm-osprey and > huwaei-taishan and the makedumpfile can work fine. > > I will come back with a follow-up patch (if needed) after some checks on > the ARMv8 Simulator for the __pte_to_phys() part. > > Thanks, > Bhupesh > > > _______________________________________________ kexec mailing list kexec@lists.infradead.org http://lists.infradead.org/mailman/listinfo/kexec