[vbox-dev] [PATCH] Fix broken SMBios checksum calculation

[Michal Necasek]

    Hi Martin,

  Now it's making sense. The '_DMI_' table header checksum code is wrong 
in DevPCBios.cpp. But dmidecode normally does not get that far, because 
it finds the '_SM_' marker first, and that one has the right checksum if 
the BIOS image got run through the 'biossums' tool. I suspect that's not 
happening for you, and that is something which is probably missing from 
our makefiles.

  I can't see into your repository on github. But you're probably right 
and it's not doing any harm. Anyway, now I see how to reproduce the 
problem, thanks.

      - Michal

On 2/5/2019 4:09 PM, Martin Fleisz wrote:
> Hi Michal,
> 
> I compile the OSE version (with some very small modifications in
> VBoxBiosAlternative386.asm
> <https://github.com/thincast/VirtualBox-RDP-Server/commit/98e7959e484a0145e39619dc6dbc40d3b9282520#diff-9c7472aa47554e70acbb1a4298b673e4>
> but that shouldn't do any harm to my understanding) and when running
> dmidecode on Ubuntu 18.04 it reports "# No SMBIOS nor DMI entry point
> found, sorry.". After applying the patch everything worked just fine.
> I checked with VirtualBox 6.0.4 and it works just fine out of the box
> ... so I guess either the checksum is somehow valid by accident or there
> is some code missing/different in the OSE repository.
> 
> Best regards, Martin
> 
> On 05.02.2019 15:48, Michal Necasek wrote:
>>
>>    Hi Martin,
>>
>>   The patch looks good, but I'm struggling to reproduce the problem.
>> The dmidecode utility in Ubuntu 18.04.1 and 18.10 (dmidecode version
>> 3.1 in both) does not generate any complaints. So what exactly does?
>>
>>       Regards,
>>         Michal
>>
>> On 2/4/2019 12:16 PM, Martin Fleisz wrote:
>>> When using legacy BIOS the Intermediate Checksum of the SMBIOS Entry
>>> Point Structure is calculated incorrectly.
>>>
>>> The calculation starts at the wrong offset (at the beginning of
>>> pu8PcBios rather than pu8PcBios[i]) and creates the sum over the
>>> whole pu8PcBios data instead of just the next 15 bytes.
>>>
>>> The patch below fixes checksum calculation and makes applications
>>> like dmidecode work again in the guest.
>>>
>>> The patch is contributed under the terms of the MIT license.
>>>
>>> Cheers, Martin
>>>
>>> diff --git src/VBox/Devices/PC/DevPcBios.cpp
>>> src/VBox/Devices/PC/DevPcBios.cpp
>>> index e967adff2f..7e56cc18fe 100644
>>> --- src/VBox/Devices/PC/DevPcBios.cpp
>>> +++ src/VBox/Devices/PC/DevPcBios.cpp
>>> @@ -1594,9 +1594,9 @@ static DECLCALLBACK(int)
>>> pcbiosConstruct(PPDMDEVINS pDevIns, int iInstance, PCF
>>>                *(uint16_t*)&pThis->pu8PcBios[i + 0x06] =
>>> RT_H2LE_U16(cbDmiTables);
>>>                *(uint16_t*)&pThis->pu8PcBios[i + 0x0C] =
>>> RT_H2LE_U16(cNumDmiTables);
>>>                uint8_t u8Sum = 0;
>>> -            for (unsigned j = 0; j < pThis->cbPcBios; j++)
>>> -                if (j != i + 0x05)
>>> -                    u8Sum += pThis->pu8PcBios[j];
>>> +            for (unsigned j = 0; j < 0x0F; j++)
>>> +                if (j != 0x05)
>>> +                    u8Sum += pThis->pu8PcBios[i + j];
>>>                pThis->pu8PcBios[i + 0x05] = -u8Sum;
>>>                break;
>>>            }
>>>
>>> _______________________________________________
>>> vbox-dev mailing list
>>> vbox-dev at virtualbox.org
>>> https://www.virtualbox.org/mailman/listinfo/vbox-dev
>>>
>>
>> _______________________________________________
>> vbox-dev mailing list
>> vbox-dev at virtualbox.org
>> https://www.virtualbox.org/mailman/listinfo/vbox-dev
> 
> 
> _______________________________________________
> vbox-dev mailing list
> vbox-dev at virtualbox.org
> https://www.virtualbox.org/mailman/listinfo/vbox-dev
>