WSL: zesInit segfaults

[tkatila]

I'd like to use Sysman API in a container in WSL. I noticed that when I call zesInit() the app segmentation faults.

I used the latest 22.04 container + latest compute-runtime release (https://github.com/intel/compute-runtime/releases/tag/24.09.28717.12).

Container is started like so: docker run -v /usr/lib/wsl:/usr/lib/wsl:ro --device /dev/dxg --rm -it ubuntu:22.04 bash

My app is essentially just a call to zesInit:

#include <zes_api.h>
#include <stdio.h>

int main()
{
        ze_result_t res = zesInit(0);
        if (res == 0) {
                fprintf(stderr, "all good\n");
        } else {
                fprintf(stderr, "all bad\n");
        }

        return 0;
}

Compiled with gcc l0.cpp -o l0 -I/usr/include/level_zero -lze_loader

GDB backtrace is like so:

 Program received signal SIGSEGV, Segmentation fault.
__strlen_avx2 () at ../sysdeps/x86_64/multiarch/strlen-avx2.S:74
74      ../sysdeps/x86_64/multiarch/strlen-avx2.S: No such file or directory.
(gdb) bt
#0  __strlen_avx2 () at ../sysdeps/x86_64/multiarch/strlen-avx2.S:74
#1  0x00007f673a926789 in std::char_traits<char>::length (__s=0x100000000 <error: Cannot access memory at address 0x100000000>)
    at /usr/include/c++/11/bits/char_traits.h:399
#2  std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::basic_string<std::allocator<char> > (__a=...,
    __s=0x100000000 <error: Cannot access memory at address 0x100000000>, this=0x56233579dba0) at /usr/include/c++/11/bits/basic_string.h:536
#3  NEO::HwDeviceIdDrm::HwDeviceIdDrm (devNodePathIn=0x0, pciPathIn=0x100000000 <error: Cannot access memory at address 0x100000000>, fileDescriptorIn=<optimized out>,
    this=0x56233579db90) at ../../neo/shared/source/os_interface/linux/hw_device_id.h:24
#4  L0::Sysman::SysmanHwDeviceIdDrm::HwDeviceIdDrm (this=0x56233579db90) at ../../neo/level_zero/sysman/source/shared/linux/sysman_hw_device_id_linux.h:17
#5  std::make_unique<L0::Sysman::SysmanHwDeviceIdDrm, int, char const*, char const*> () at /usr/include/c++/11/bits/unique_ptr.h:962
#6  L0::Sysman::createSysmanHwDeviceId (hwDeviceId=std::unique_ptr<NEO::HwDeviceId> = {...})
    at ../../neo/level_zero/sysman/source/shared/linux/sysman_hw_device_id_linux.cpp:48
#7  0x00007f673a91e0b6 in L0::Sysman::SysmanDriverImp::initialize (this=<optimized out>, result=0x7ffc20c4c504)
    at ../../neo/level_zero/sysman/source/driver/sysman_driver.cpp:49
#8  0x00007f673a91de8f in operator() (__closure=<optimized out>) at ../../neo/level_zero/sysman/source/driver/sysman_driver.cpp:78
#9  std::__invoke_impl<void, L0::Sysman::SysmanDriverImp::driverInit(zes_init_flags_t)::<lambda()> > (__f=...) at /usr/include/c++/11/bits/invoke.h:61
#10 std::__invoke<L0::Sysman::SysmanDriverImp::driverInit(zes_init_flags_t)::<lambda()> > (__fn=...) at /usr/include/c++/11/bits/invoke.h:96
#11 operator() (__closure=<optimized out>) at /usr/include/c++/11/mutex:776
#12 operator() (__closure=0x0) at /usr/include/c++/11/mutex:712
#13 _FUN () at /usr/include/c++/11/mutex:712
#14 0x00007f673c862ee8 in __pthread_once_slow (once_control=0x7f673c3f6aa8 <L0::Sysman::driverImp+8>, init_routine=0x7f673c677d50 <__once_proxy>)
    at ./nptl/pthread_once.c:116
#15 0x00007f673a91e589 in __gthread_once (__func=<optimized out>, __once=0x7f673c3f6aa8 <L0::Sysman::driverImp+8>)
    at /usr/include/x86_64-linux-gnu/c++/11/bits/gthr-default.h:700
#16 std::call_once<L0::Sysman::SysmanDriverImp::driverInit(zes_init_flags_t)::<lambda()> > (__f=..., __once=...) at /usr/include/c++/11/mutex:783
#17 L0::Sysman::SysmanDriverImp::driverInit (flags=<optimized out>, this=0x7f673c3f6aa0 <L0::Sysman::driverImp>)
    at ../../neo/level_zero/sysman/source/driver/sysman_driver.cpp:76
#18 L0::Sysman::init (flags=<optimized out>) at ../../neo/level_zero/sysman/source/driver/sysman_driver.cpp:119
#19 0x00007f673ca27b5b in zesInit () from /lib/x86_64-linux-gnu/libze_loader.so.1
#20 0x0000562332e7617f in main () at l0.cpp:6

[saik-intel]

We will look into this issue , ideally it shouldn't segfault

[eero-t]

We will look into this issue , ideally it shouldn't segfault

@saik-intel / @JablonskiMateusz (nearly) 4 months gone, any news?

[saik-intel]

we didn't fixed yet, will work on lower priority.