crash on AMD

[oltolm]

It crashes on AMD.

cpufetch.exe caused an Unknown [0xC0000374] Exception at location 00007FFC021AF349 in module ntdll.dll.

AddrPC           Params
00007FFC021AF349 00007FFC021F762C 0000007897DFFAF0 0000026ED8861560  ntdll.dll!RtlIsZeroMemory+0x119
00007FFC021AF313 00007FFC005FFA30 00007FFC022197F0 0000000000000009  ntdll.dll!RtlIsZeroMemory+0xe3
00007FFC021B8092 0000000000000009 0000000000000000 0000026ED8860000  ntdll.dll!RtlpNtSetValueKey+0x4b2
00007FFC021B837A 0000026ED8860000 0000026ED8860000 0000000000000000  ntdll.dll!RtlpNtSetValueKey+0x79a
00007FFC021BE001 0000000000000074 0000026ED8861560 0000026ED8861570  ntdll.dll!RtlpNtSetValueKey+0x6421
00007FFC020D5B43 0000000000000074 0000026ED8860000 0000026ED88677D0  ntdll.dll!RtlGetCurrentServiceSessionId+0x12f3
00007FFC020D47B1 0000026ED8861660 0000000000000013 0000000000000000  ntdll.dll!RtlFreeHeap+0x51
00007FFC00589C9C 0000026ED88617F8 0000000000000013 0000026ED88616F0  msvcrt.dll!free+0x1c
00007FF6C5207785 0000000000000000 0000026ED8861410 003600D100000060  cpufetch.exe!free_topo_struct+0x15  [C:/src/cpufetch/src/x86/cpuid.c @ 1173]
  1171: 
  1172: void free_topo_struct(struct topology* topo) {
> 1173:   free(topo->apic->cache_select_mask);
  1174:   free(topo->apic->cache_id_apic);
  1175:   free(topo->apic);
00007FF6C520429E 0000026ED8861410 0000000000000000 0000000000000000  cpufetch.exe!print_cpufetch_x86+0x68e  [C:/src/cpufetch/src/common/printer.c @ 653]
   651:   if(cs != NULL) free_colors_struct(cs);
   652:   if(cpu->cach != NULL) free_cache_struct(cpu->cach);
>  653:   if(cpu->topo != NULL) free_topo_struct(cpu->topo);
   654:   free_freq_struct(cpu->freq);
   655:   free_cpuinfo_struct(cpu);
00007FF6C520463F 0000000000000000 00007FF6C5201419 0000000000000008  cpufetch.exe!print_cpufetch+0x2f  [C:/src/cpufetch/src/common/printer.c @ 1090]
  1088: 
  1089: #ifdef ARCH_X86
> 1090:   return print_cpufetch_x86(cpu, s, cs, term, show_full_cpu_name);
  1091: #elif ARCH_PPC
  1092:   return print_cpufetch_ppc(cpu, s, cs, term, show_full_cpu_name);
00007FF6C521A2CF 0000000000000000 000000000000000D 00007FF6C522B308  cpufetch.exe!main+0x8f  [C:/src/cpufetch/src/common/main.c @ 123]
   121:   }
   122: 
>  123:   if(print_cpufetch(cpu, get_style(), get_colors(), show_full_cpu_name())) {
   124:     return EXIT_SUCCESS;
   125:   }
00007FF6C52012EE 0000000000000000 0000000000000000 0000000000000000  cpufetch.exe!__tmainCRTStartup+0x16e  [C:/M/B/src/mingw-w64/mingw-w64-crt/crt/crtexe.c @ 267]
00007FF6C5201406 0000000000000000 0000000000000000 0000000000000000  cpufetch.exe!mainCRTStartup+0x16  [C:/M/B/src/mingw-w64/mingw-w64-crt/crt/crtexe.c @ 188]
00007FFC00357344 0000000000000000 0000000000000000 0000000000000000  KERNEL32.DLL!BaseThreadInitThunk+0x14
00007FFC021026B1 0000000000000000 0000000000000000 0000000000000000  ntdll.dll!RtlUserThreadStart+0x21

[Dr-Noob]

Well, I'm amazed by the fact that this issue has not been discovered before because this bug affects all AMD CPUs, although it seems like it does not generally cause the segfault.

I've pushed a commit in the i212 branch. Can you confirm that it fixes the issue?

[oltolm]

Yes it fixes the issue. BTW you don't need to check for NULL before calling free. free does nothing if the argument is NULL.

[Dr-Noob]

Yes it fixes the issue. BTW you don't need to check for NULL before calling free. free does nothing if the argument is NULL.

You are right. The fix actually comes from calling calloc which will set the pointer to NULL so that free won´t try to free a random memory address. I'll have a look at this, since there are other places where NULL is also checked before calling free (and, possibly, the right thing to do is actually to have a calloc call, not to check for NULL).