[ARC64][specint2006] omnetpp segfaults

[abrodkin]

Fails exactly the same in both QEMU & nSIM

1. nSIM: nsimdrv -prop=nsim_isa_vec64=1 -prop=nsim_isa_family=arc64 -prop=nsim_isa_enable_timer_0=1 -prop=nsim_isa_enable_timer_1=1 -prop=nsim_isa_addr_size=64 -prop=nsim_isa_pc_size=64 -prop=icache=16384,64,4,o -prop=dcache=16384,64,4,o -prop=mmu_version=6 -prop=mmu_pagesize=4096 -prop=mmu_address_space=48 -prop=nsim_mem-dev=uart0,kind=dwuart,base=0xf0000000,irq=24 -prop=nsim_isa_mpy_option=3 -prop=nsim_isa_mpy64=1 -prop=nsim_isa_div64_option=1 -prop=nsim_isa_div_rem_option=2 -prop=nsim_isa_atomic_option=1 ./vmlinux
2. QEMU: qemu-system-arc64 -M simhs -global cpu.mmuv6=true -nographic -no-reboot -serial stdio -monitor none -cpu arc64 -kernel vmlinux

# cd /arc64-spec-test/benchmarks/471.omnetpp/
# ./omnetpp omnetpp.ini
OMNeT++ Discrete Event Simulation  (C) 1992-2003 Andras Varga
See the license for distribution terms and warranty disclaimer
Setting up Cmdenv (command-line user interface)...

Preparing for Run #1...
Setting up network `twoHosts'...
random: fast init done
Running simulation...
twoHosts.hostA.srv registering DSAP 241
twoHosts.hostB.srv registering DSAP 241
twoHosts.hostA.cli registering DSAP 240
twoHosts.hostB.cli registering DSAP 240
** Event #0   T=0.0000000  ( 0.00s)
     Messages:  created: 20
** Event #100000   T=14.836799  (14.83s)
     Messages:  created: 48540
** Event #200000   T=29.571801  (29.57s)
     Messages:  created: 97064
** Event #300000   T=44.401886  (44.40s)
     Messages:  created: 145570
** Event #400000   T=59.434241  (59.43s)
     Messages:  created: 194096
** Event #500000   T=74.211485 ( 1m 14s)
     Messages:  created: 242608
** Event #600000   T=89.306257 ( 1m 29s)
     Messages:  created: 291116
** Event #700000   T=103.92283 ( 1m 43s)
     Messages:  created: 339640
** Event #800000   T=119.27061 ( 1m 59s)
     Messages:  created: 388154
** Event #900000   T=133.81414 ( 2m 13s)
     Messages:  created: 436666
** Event #1000000   T=148.38048 ( 2m 28s)
     Messages:  created: 485196
** Event #1100000   T=163.10847 ( 2m 43s)
     Messages:  created: 533702
** Event #1200000   T=177.83484 ( 2m 57s)
     Messages:  created: 582219
potentially unexpected fatal signal 11.
Path: /arc64-spec-test/benchmarks/471.omnetpp/omnetpp
CPU: 0 PID: 68 Comm: omnetpp Not tainted 5.6.0 #7
Invalid Write @ 0x00000000 by insn @ 0x202312c2
  @off 0xa2c2 in [/lib/libgcc_s.so.1]  VMA: 0x20227000 to 0x20237000
ECR: 0x00050200 EFA: 0x00000000 ERET: 0x202312c2
STAT32: 0x80081882 [IE U     ]   BTA: 0x200cb2e4
 SP: 0x5fffe438  FP: 0x00000000 BLK: 0x200cb61e
r00: 0x00000000 r01: 0x00000000 r02: 0x007b3888
r03: 0x00000000 r04: 0x00000008 r05: 0x00013620
r06: 0x0000007c r07: 0x00000000 r08: 0x202390f0
r09: 0x000467f3 r10: 0x20012f74 r11: 0x20021170
r12: 0x0001362c r13: 0xfefefefefefefeff r14: 0x20231274
r15: 0x007b3888 r16: 0x5fffefd0 r17: 0x00046852
r18: 0x00000000 r19: 0x00000000 r20: 0x000a9350
r21: 0x00000001 r22: 0x00000001 r23: 0x00000001
r24: 0x00000000 r25: 0x00000020
Segmentation fault

Attached vmlinux-arc64-specint2006-omnetpp.2021.04.01.zip

To extract initramfs.cpio.gz do arc64-elf-objcopy -O binary --only-section=.init.ramfs vmlinux initramfs.cpio.gz.

[abrodkin]

That's the part of libgcc_s.so.1 where it fails:

000000000000a274 <_Unwind_SetGR@@GCC_3.0>:
    a274:»      218c 88c1           »   cmp»    r1,99
    a278:»      f6c7                »   ble_s»  14»     ;a286 <_Unwind_SetGR@@GCC_3.0+0x12>
    a27a:»      c7e7                »   pushl_s»blink
    a27c:»      2731 ff00 0000 7bcc »   ldl»    r0,[pcl,31692@s32]»     ;11e48 <__emutls_register_common@@GCC_4.3.0+0x43b8>
    a284:»      7840                »   jl_s»   [r0]
    a286:»      592f 0047           »   sexwl»  r1,r1
    a28a:»      5f00 7f03 0000 7df4 »   addl»   r3,pcl,32244@s32»       ;1207c <__emutls_register_common@@GCC_4.3.0+0x45ec>
    a292:»      2332 0044           »   ldb»    r4,[r3,r1]
    a296:»      106a 0e43           »   ldl.as» r3,[r0,106]
    a29a:»      5b51 8f80           »   btstl»  r3,0x3e
    a29e:»      f20e                »   beq_s»  28»     ;a2b8 <_Unwind_SetGR@@GCC_3.0+0x44>
    a2a0:»      5800 0043           »   addl»   r3,r0,r1
    a2a4:»      5b96 0b41           »   add3l»  r3,r3,109
    a2a8:»      8b60                »   ldb_s»  r3,[r3,0]
    a2aa:»      7b6b                »   tst_s»  r3,r3
    a2ac:»      f206                »   beq_s»  12»     ;a2b8 <_Unwind_SetGR@@GCC_3.0+0x44>
    a2ae:»      5816 0041           »   add3l»  r1,r0,r1
    a2b2:»      1900 0087           »   stl»    r2,[r1]
    a2b6:»      7ee0                »   j_s»    [blink]
    a2b8:»      20f1 8040           »   ldl.as» r0,[r0,r1]
    a2bc:»      244c 8200           »   cmp»    r4,0x8
    a2c0:»      f5dd                »   bne_s»  -70»    ;a27a <_Unwind_SetGR@@GCC_3.0+0x6>
    a2c2:»      1800 0087           »   stl»    r2,[r0]   <--- HERE WE FAIL
    a2c6:»      7ee0                »   j_s»    [blink]

@claziss @vineetgarc yet another unwinder-related problem? Or #38 is only applicable to the kernel and user-space unwinder is supposed to work?

[claziss]

It is libgcc, it is unwind procedure, probably it is unwinding issue.