UnwindManagedExceptionPass1 is failing on illumos amd64

[am11]

With a simple app that throws exception:

using System;
class Program
{
    static void Main(string[] args)
    {
        if (Array.IndexOf(args, "--throw") > -1)
            try
            {
                throw new Exception("Hello exception!");
            }
            catch (Exception ex)
            {
               Console.WriteLine(ex.Message);
               throw;
            }
...

after the exception is rethrown and message is printed to stderr, application abort()s on SmartOS amd64.

$ LD_PRELOAD=/lib/amd64/libumem.so ~/.dotnet/dotnet test-app/bin/Debug/net5.0/publish/test-app.dll --throw

Hello exception!
Unhandled exception. System.Exception: Hello exception!
   at test-app.Program.Main(String[] args)
Abort (core dumped)

On Linux, it cleanly exits the process (without abort()).

coredump inspection yields:

$ sudo mdb /var/cores/dotnet.df333bd6-c7f8-63b6-b1c2-e947d740e084.6546.1593009420.core 
mdb: core file data for mapping at fffffc7f4d822000 not saved: Bad address
mdb: core file data for mapping at fffffc7f5d812000 not saved: Bad address
mdb: core file data for mapping at fffffc7f65802000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d810000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d851000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d85e000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d861000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d867000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d871000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d875000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d8a6000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d978000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6d9ee000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6dc6f000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6de7d000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6df59000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6e00c000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6e08f000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6e0e8000 not saved: Bad address
mdb: core file data for mapping at fffffc7f6e0f6000 not saved: Bad address
mdb: core file data for mapping at fffffc7fed8a1000 not saved: Bad address
mdb: core file data for mapping at fffffc7fee1a0000 not saved: Bad address
mdb: core file data for mapping at fffffc7fee490000 not saved: Bad address
mdb: core file data for mapping at fffffc7fee6a0000 not saved: Bad address
mdb: core file data for mapping at fffffc7fef000000 not saved: Bad address
mdb: core file data for mapping at fffffc7fef00f000 not saved: Bad address
mdb: core file data for mapping at fffffc7fef012000 not saved: Bad address
mdb: core file data for mapping at fffffc7fef040000 not saved: Bad address
mdb: core file data for mapping at fffffc7fef050000 not saved: Bad address
Loading modules: [ libumem.so.1 libc.so.1 ld.so.1 ]
> ::status
debugging core file of dotnet (64-bit) from df333bd6-c7f8-63b6-b1c2-e947d740e084
file: /home/am11/.dotnet/dotnet
initial argv: /home/am11/.dotnet/dotnet test-app/bin/Debug/net5.0/publish/test-app.dll
threading model: native threads
status: process terminated by SIGABRT (Abort), pid=6546 uid=1000 code=-1
> $C
fffffc7fffdfbed0 libc.so.1`_lwp_kill+0xa()
fffffc7fffdfbf00 libc.so.1`raise+0x1e(6)
fffffc7fffdfbf50 libc.so.1`abort+0x58()
fffffc7fffdfbf60 libcoreclr.so`InitializeFlushProcessWriteBuffers()
fffffc7fffdfbfc0 libcoreclr.so`_ZL14PROCEndProcessPvji+0x3b1()
fffffc7fffdfc000 libcoreclr.so`TerminateProcess+0xbd()
fffffc7fffdfc020 libcoreclr.so`_Z28CrashDumpAndTerminateProcessj+0x20()
fffffc7fffdfc6c0 libcoreclr.so`_Z27UnwindManagedExceptionPass1R16PAL_SEHExceptionP8_CONTEXT+0x628()
fffffc7fffdfcc30 libcoreclr.so`_Z24DispatchManagedExceptionR16PAL_SEHExceptionb+0x110()
fffffc7fffdfcde0 libcoreclr.so`_Z10IL_Rethrowv+0x1f0()
fffffc7fffdfe5f0 0xfffffc7f6d9ca14f()
fffffc7fffdfe610 libcoreclr.so`CallDescrWorkerInternal+0x7c()
fffffc7fffdfe680 libcoreclr.so`_Z26CallDescrWorkerWithHandlerP13CallDescrDatai+0x11b()
fffffc7fffdfe8c0 libcoreclr.so`_ZN18MethodDescCallSite16CallTargetWorkerEPKmPmi+0x8ef()
fffffc7fffdfe8e0 libcoreclr.so`_ZN18MethodDescCallSite4CallEPKm+0x2a()
fffffc7fffdfea60 libcoreclr.so`_ZL15RunMainInternalP5Param+0x26b()
fffffc7fffdfea80 libcoreclr.so`_ZZZ7RunMainP10MethodDescsPiP3REFI8PtrArrayEENKUlP5ParamE_clES7_ENKUlS7_E_clES7_+0x1c()
fffffc7fffdfeb00 libcoreclr.so`_ZZ7RunMainP10MethodDescsPiP3REFI8PtrArrayEENKUlP5ParamE_clES7_+0x5f()
fffffc7fffdfebe0 libcoreclr.so`_Z7RunMainP10MethodDescsPiP3REFI8PtrArrayE+0x1db()
fffffc7fffdfee90 libcoreclr.so`_ZN8Assembly17ExecuteMainMethodEP3REFI8PtrArrayEi+0x113()
fffffc7fffdff010 libcoreclr.so`_ZN8CorHost215ExecuteAssemblyEjPKDsiPS1_Pj+0x3ff()
fffffc7fffdff0d0 libcoreclr.so`coreclr_execute_assembly+0x127()
fffffc7fffdff110 libhostpolicy.so`_ZN9coreclr_t16execute_assemblyEiPPKcS1_Pj+0x7c()
fffffc7fffdff250 libhostpolicy.so`_Z19run_app_for_contextRK20hostpolicy_context_tiPPKc+0x44b()
fffffc7fffdff2a0 libhostpolicy.so`_Z7run_appiPPKc+0x6d()
fffffc7fffdff450 libhostpolicy.so`corehost_main+0x15b()
fffffc7fffdff540 libhostfxr.so`_ZL11execute_appRKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEP15corehost_init_tiPPKc+0x2f2()
fffffc7fffdff5e0 libhostfxr.so`_ZN12_GLOBAL__N_123read_config_and_executeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEERK19host_startup_info_tS7_RKSt13unordered_mapI13known_optionsSt6vectorIS5_SaIS5_EE18known_options_hashSt8equal_toISC_ESaISt4pairIKSC_SF_EEEiPPKc11host_mode_tPciPi+0xd7()
fffffc7fffdff6a0 libhostfxr.so`_ZN10fx_muxer_t24handle_exec_host_commandERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEERK19host_startup_info_tS7_RKSt13unordered_mapI13known_optionsSt6vectorIS5_SaIS5_EE18known_options_hashSt8equal_toISC_ESaISt4pairIKSC_SF_EEEiPPKci11host_mode_tPciPi+0x175()
fffffc7fffdff7a0 libhostfxr.so`_ZN10fx_muxer_t7executeENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEiPPKcRK19host_startup_info_tPciPi+0x166()
fffffc7fffdff890 libhostfxr.so`hostfxr_main_startupinfo+0xac()
fffffc7fffdffa30 _Z9exe_startiPPKc+0x49c()
fffffc7fffdffa60 main+0xa5()
fffffc7fffdffa90 _start_crt+0x83()
fffffc7fffdffaa0 _start+0x18()

From libunwind, we hit this condition:

runtime/src/coreclr/src/pal/src/libunwind/src/x86_64/Gstep.c

Line 276 in 581dc19

Debug (2, "returning 0\n");

then in UnwindManagedExceptionPass1, since we are at IP==0, we get here:

runtime/src/coreclr/src/vm/exceptionhandling.cpp

Line 4615 in 581dc19

CrashDumpAndTerminateProcess(1);

stress trace looks like this:

Hello exception!
Processing exception at native frame: IP = FFFFFC7FE73925C4, SP = FFFFFC7FFFDFCE40 
Processing exception at native frame: IP = FFFFFC7FE7393463, SP = FFFFFC7FFFDFCED0 
Processing exception at native frame: IP = FFFFFC7FE7399D87, SP = FFFFFC7FFFDFCFA0 
Processing exception at native frame: IP = FFFFFC7FE739A369, SP = FFFFFC7FFFDFD1B0 
Processing exception at native frame: IP = FFFFFC7FE739AAD0, SP = FFFFFC7FFFDFD7E0 
Processing exception at native frame: IP = FFFFFC7FE72AE791, SP = FFFFFC7FFFDFDE80 
Processing exception at native frame: IP = FFFFFC7F6D5CA105, SP = FFFFFC7FFFDFE3F0 
Processing exception at native frame: IP = FFFFFC7FE7234446, SP = FFFFFC7FFFDFE630 
Processing exception at native frame: IP = FFFFFC7FE72350B7, SP = FFFFFC7FFFDFE650 
Processing exception at native frame: IP = FFFFFC7FE7108414, SP = FFFFFC7FFFDFE6C0 
Processing exception at native frame: IP = FFFFFC7FE73CAA96, SP = FFFFFC7FFFDFE900 
Processing exception at native frame: IP = FFFFFC7FE73CAB64, SP = FFFFFC7FFFDFE920 
Processing exception at native frame: IP = FFFFFC7FE73CABE5, SP = FFFFFC7FFFDFEAA0 
Processing exception at native frame: IP = FFFFFC7FE73CAEAA, SP = FFFFFC7FFFDFEAC0 
Processing exception at native frame: IP = FFFFFC7FE73CB1ED, SP = FFFFFC7FFFDFEB40 
Processing exception at native frame: IP = FFFFFC7FE7103E11, SP = FFFFFC7FFFDFEC20 
Processing exception at native frame: IP = FFFFFC7FE70DF4F1, SP = FFFFFC7FFFDFEED0 
Processing exception at native frame: IP = FFFFFC7FE8222EE4, SP = FFFFFC7FFFDFF050 
Processing exception at native frame: IP = FFFFFC7FE8238FA8, SP = FFFFFC7FFFDFF110 
Processing exception at native frame: IP = FFFFFC7FE823921D, SP = FFFFFC7FFFDFF150 
Processing exception at native frame: IP = FFFFFC7FE8239846, SP = FFFFFC7FFFDFF290 
Processing exception at native frame: IP = FFFFFC7FE862D84C, SP = FFFFFC7FFFDFF2E0 
Processing exception at native frame: IP = FFFFFC7FE862F450, SP = FFFFFC7FFFDFF490 
Processing exception at native frame: IP = FFFFFC7FE8630FB9, SP = FFFFFC7FFFDFF580 
Processing exception at native frame: IP = FFFFFC7FE862F662, SP = FFFFFC7FFFDFF620 
Processing exception at native frame: IP = FFFFFC7FE862ABD9, SP = FFFFFC7FFFDFF6E0 
Processing exception at native frame: IP = 000000000040E33B, SP = FFFFFC7FFFDFF7E0 
Processing exception at native frame: IP = 000000000040E625, SP = FFFFFC7FFFDFF8D0 
Processing exception at native frame: IP = 00000000004031D3, SP = FFFFFC7FFFDFFA70 
Processing exception at native frame: IP = 0000000000403138, SP = FFFFFC7FFFDFFAA0 
returning 0     <-- this one from libunwind 
Processing exception at native frame: IP = 0000000000000000, SP = FFFFFC7FFFDFFAD0
Unhandled exception. System.Exception: Hello exception!
   at test-app.Program.Main(String[] args)
Abort (core dumped)

[Dotnet-GitSync-Bot]

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[am11]

@jasonbking suggested that if it is libunwind misbehaving, it would be helpful to come up with a short repro (preferably in plain C).

@janvorli, is there any such repro available offhand? I am not sure whether to cut off the assembly stubs (RtlCaptureContext and friends) in repro.

[janvorli]

The behavior you are seeing is correct. Unhandled exception should exit via abort and it does so on Linux too. This is from my Ubuntu 16.04:

 ~/dotnet50/dotnet /home/janvorli/test/testunhandled/bin/Debug/net5.0/testunhandled.dll --throw
Hello exception!
Unhandled exception. System.Exception: Hello exception!
   at Program.Main(String[] args) in /home/janvorli/test/testunhandled/Program.cs:line 9
Aborted (core dumped)

[am11]

I was testing on Ubuntu 18.04. Although the exit code was 134, wasn't giving Aborted (core dumped). This message comes from shell when an app (bash job) terminates from a signal. e.g. a simple C program produces that in the same shell session:

gcc -xc - <<EOF
#include <stdlib.h>
int main(void)
{
    abort();
    return (0);
}
EOF
./a.out
Aborted (core dumped)

[am11]

Ah, I was doing dotnet run --throw which fork()s, instead of dotnet path/to/file.dll --throw. So it makes sense. :)