Re: [Crash-utility] crash tool fail to analyze the image

----- Original Message ----- Very interesting. The support of arm64 is completely theoretical -- I wrote it all based upon kernel sources with absolutely no testing whatsoever. I have been waiting for full access to arm64 hardware, and it appears I should have that access within a few weeks. Right -- that warning is a placeholder. You could comment out that warning message for now. The "read error" messages indicate that the physical addresses that were calculated from the virtual addresses were not found in any of the PT_LOAD segments in the ELF header. I'm guessing that the "phys_offset" calculation returned by arm64_calc_phys_offset() is maybe causing the arm64 VTOP() to come up with bogus values: #define VTOP(X) \ ((unsigned long)(X)-(machdep->kvbase)+(machdep->machspec->phys_offset)) machdep_init(PRE_GDB) arm64_init(PRE_GDB) arm64_calc_phys_offset() arm64_kdump_phys_base() arm_kdump_phys_base() But again, that's just a guess. If you turn on debugging with "crash -d<number> ...", you might see more useful information. Right -- try "crash -d1 ..." or "crash -d8 ..." for more info. The larger the number, the more verbose the debug output will be. But the first question is: how did you create the vmcore? Are you working on implementing support for arm64 kexec/kdump support in the kernel? Yes, I would be very interested if you could make the vmlinux/vmcore pair available to me. I will send you information on how to do that offline. Thanks, Dave