On 19.10.20 23:36, Nick Desaulniers wrote: > On Mon, Oct 19, 2020 at 3:55 AM Christian Borntraeger > <borntraeger@xxxxxxxxxx> wrote: >> >> >> >> On 17.10.20 01:40, Nick Desaulniers wrote: >>> On Fri, Oct 16, 2020 at 4:18 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote: >>>> >>>> On 10/16/20 4:11 PM, Nick Desaulniers wrote: >>>>> Hello all, >>>>> I'm working on integrating the latest release of buildroot (2020.08.1) >>>>> into our CI for ClangBuiltLinux. >>>>> >>>>> https://github.com/ClangBuiltLinux/boot-utils/pull/25 >>>>> https://github.com/ClangBuiltLinux/boot-utils/pull/26 >>>>> https://github.com/ClangBuiltLinux/continuous-integration/pull/327 >>>>> >>>>> I'm seeing the following error from QEMU: >>>>> KASLR disabled: CPU has no PRNG >>>>> Linux version 5.9.0-00732-g04ed4527465f (ndesaulniers@<myhost>) #30 >>>>> SMP Fri Oct 16 15:49:05 PDT 2020Kernel fault: interruption code 0005 >>>>> ilc:2 >>>>> PSW : 0000200180000000 000000000001779e >>>>> R:0 T:0 IO:0 EX:0 Key:0 M:0 W:0 P:0 AS:0 CC:2 PM:0 RI:0 EA:3 >>>>> GPRS: 0000000000000001 0000000c00000000 00000003fffffff4 00000000fffffff0 >>>>> 0000000000000000 00000000fffffff4 000000000000000c 00000000fffffff0 >>>>> 00000000fffffffc 0000000000000000 00000000fffffff8 00000000008a75a8 >>>>> 0000000000000009 0000000000000002 0000000000000008 000000000000bce0 >> >> Do you have more information? If not, any chance to run objdump on the kernel >> and check what instructions do you have aroung address 0x1779e >> [...] > > Is PSW the program counter? program counter (the 64bit on the right hand side), flags, interrupt controls and many other things > /me skims https://www.kernel.org/doc/Documentation/s390/Debugging390.txt *woah* > Is there something more specific I should be running than: > $ s390x-linux-gnu-objdump -Dr vmlinux | grep 179fe > because I get lots of hits for that, but not 0x179fe. You probably have to look at the instruction before as many faults on s390 are suppressing (the instruction effects are suppressed, but the PSW moves forward). With ilc=2 (instruction length code) > > In fact, the first symbol starts at 0x100000. Unless this is a failure > in the image decompressor? I guess so. Maybe llvm does something that qemu does not handle in the compressor can you try with objdump -d arch/s390/boot/compressed/vmlinux
