On 05/23/19 10:37, Peter Maydell wrote: > On Thu, 23 May 2019 at 01:51, Laszlo Ersek <ler...@redhat.com> wrote: >> I have narrowed down the issue sufficiently that I think I can hand >> it over to Peter and Ard now -- because they know AARCH32 and AARCH64 >> assembly, and "target/arm/translate-a64.c" and "tcg/arm/*" too. >> >> The summarize the issue for Ard, the symptom is that AARCH64 >> ArmVirtQemu runs perfectly fine with TCG on an x86-64 system, but it >> crashes on an AARCH32 host system. > > Thanks for the investigation; this is probably more one for Richard > than me.
I figured I'd provide some logs. (1) To recap, the aarch64 assembly source code file that seems to be mis-translated (from the aarch64 binary) lives at https://github.com/tianocore/edk2/blob/3604174718e2afc950c3cc64c64ba5165c8692bd/MdePkg/Library/BaseMemoryLibOptDxe/AArch64/CopyMem.S The relevant functions are InternalMemCopyMem() and __memcpy(). (2) I disassembled the aarch64 binary with "objdump", and uploaded the listing to: http://people.redhat.com/lersek/aarch64-to-arm-mistranslation/DxeCore.objdump.xz The InternalMemCopyMem() function starts at hex offset 2b2ec. The __memcpy() function starts at hex offset 2b180. (3) I ran the guest with "-d in_asm,op,op_opt,op_ind,out_asm" and uploaded the log file to: http://people.redhat.com/lersek/aarch64-to-arm-mistranslation/tcg.in_asm.op.op_opt.op_ind.out_asm.log.xz The TBs that correspond to (parts of) the InternalMemCopyMem() and __memcpy() functions are scattered over the file, but the offset between the "nice" disassembly from (2), and the in-RAM TBs in (3), can be determined from the fact that there is a single prfm instruction in the entire binary. The instruction's offset is 0x2b180 in (2) -- at the beginning of the __memcpy() function --, and its RAM address is 0x472d2180 in (3). Thus the difference (= the load address of DxeCore.efi) is 0x472a7000. (This is logged by the guest as well: > Loading DXE CORE at 0x000472A7000 EntryPoint=0x000472A8000 ) Thanks Laszlo
