block: Fix vpc initialization of the Dynamic Disk Header

The Data Offset field in the Dynamic Disk Header is an 8 byte field.
Although the specification (2006-10-11) gives an example of initializing
only the first 4 bytes, images generated by Microsoft on Windows initialize 
all 8 bytes.

Failure to initialize all 8 bytes results in errors from utilities
that check specifically for the complete Data Offset field initialization.

Signed-off-by: Charles Arnold <carnold@suse.com>

Am 08.11.2011 20:25, schrieb Charles Arnold:
> The Data Offset field in the Dynamic Disk Header is an 8 byte field.
> Although the specification (2006-10-11) gives an example of initializing
> only the first 4 bytes, images generated by Microsoft on Windows initialize 
> all 8 bytes.
> 
> Failure to initialize all 8 bytes results in errors from utilities
> that check specifically for the complete Data Offset field initialization.
> 
> Signed-off-by: Charles Arnold <carnold@suse.com>

Reviewed-by: Andreas Färber <afaerber@suse.de>

Kevin, can you please apply this for 1.0? Thanks!

Andreas

> 
> diff --git a/block/vpc.c b/block/vpc.c
> index 416f489..35ac3fd 100644
> --- a/block/vpc.c
> +++ b/block/vpc.c
> @@ -585,7 +585,7 @@ static int vpc_create(const char *filename, QEMUOptionParameter *options)
> 
>      memcpy(dyndisk_header->magic, "cxsparse", 8);
> 
> -    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
> +    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
>      dyndisk_header->table_offset = be64_to_cpu(3 * 512);
>      dyndisk_header->version = be32_to_cpu(0x00010000);
>      dyndisk_header->block_size = be32_to_cpu(block_size);

Am 08.11.2011 20:25, schrieb Charles Arnold:
> The Data Offset field in the Dynamic Disk Header is an 8 byte field.
> Although the specification (2006-10-11) gives an example of initializing
> only the first 4 bytes, images generated by Microsoft on Windows initialize 
> all 8 bytes.
> 
> Failure to initialize all 8 bytes results in errors from utilities
> that check specifically for the complete Data Offset field initialization.
> 
> Signed-off-by: Charles Arnold <carnold@suse.com>
> 
> diff --git a/block/vpc.c b/block/vpc.c
> index 416f489..35ac3fd 100644
> --- a/block/vpc.c
> +++ b/block/vpc.c
> @@ -585,7 +585,7 @@ static int vpc_create(const char *filename, QEMUOptionParameter *options)
> 
>      memcpy(dyndisk_header->magic, "cxsparse", 8);
> 
> -    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
> +    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
>      dyndisk_header->table_offset = be64_to_cpu(3 * 512);
>      dyndisk_header->version = be32_to_cpu(0x00010000);
>      dyndisk_header->block_size = be32_to_cpu(block_size);

Can you please add a short comment explaining why we deviate from the
specification in this point? If someone notices the discrepancy later,
he shouldn't have to dig up the git commit message.

Kevin

>>> On 11/9/2011 at 02:46 AM, in message <4EBA4BEC.5040401@redhat.com>, Kevin Wolf
<kwolf@redhat.com> wrote: 
> Am 08.11.2011 20:25, schrieb Charles Arnold:
>> The Data Offset field in the Dynamic Disk Header is an 8 byte field.
>> Although the specification (2006-10-11) gives an example of initializing
>> only the first 4 bytes, images generated by Microsoft on Windows initialize 
>> all 8 bytes.
>> 
>> Failure to initialize all 8 bytes results in errors from utilities
>> that check specifically for the complete Data Offset field initialization.
>> 
>> Signed-off-by: Charles Arnold <carnold@suse.com>
>> 
>> diff --git a/block/vpc.c b/block/vpc.c
>> index 416f489..35ac3fd 100644
>> --- a/block/vpc.c
>> +++ b/block/vpc.c
>> @@ -585,7 +585,7 @@ static int vpc_create(const char *filename, 
> QEMUOptionParameter *options)
>> 
>>      memcpy(dyndisk_header->magic, "cxsparse", 8);
>> 
>> -    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
>> +    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
>>      dyndisk_header->table_offset = be64_to_cpu(3 * 512);
>>      dyndisk_header->version = be32_to_cpu(0x00010000);
>>      dyndisk_header->block_size = be32_to_cpu(block_size);
> 
> Can you please add a short comment explaining why we deviate from the
> specification in this point? If someone notices the discrepancy later,
> he shouldn't have to dig up the git commit message.
> 
> Kevin
 
Using qemu-img to convert a raw image to a vhd image seems to work fine accept that only 4 bytes of the 8 byte Dynamic Disk Header Data Offset field are initialized.  This seems correct according to the specification but tools like Citrix's vhd-util fail when querying the image due to the shortened initialization.  Upon further analysis, it was determined that Microsoft also initializes all 8 bytes when generating a vhd image on windows.  Initializing all 8 bytes seems more logical and may indicate that Microsoft needs to update the specification because the requirement for this field changed or that it is simply an oversight that has never been noticed.

- Charles

Am 09.11.2011 16:42, schrieb Charles Arnold:
>>>> On 11/9/2011 at 02:46 AM, in message <4EBA4BEC.5040401@redhat.com>, Kevin Wolf
> <kwolf@redhat.com> wrote: 
>> Am 08.11.2011 20:25, schrieb Charles Arnold:
>>> The Data Offset field in the Dynamic Disk Header is an 8 byte field.
>>> Although the specification (2006-10-11) gives an example of initializing
>>> only the first 4 bytes, images generated by Microsoft on Windows initialize 
>>> all 8 bytes.
>>>
>>> Failure to initialize all 8 bytes results in errors from utilities
>>> that check specifically for the complete Data Offset field initialization.
>>>
>>> Signed-off-by: Charles Arnold <carnold@suse.com>
>>>
>>> diff --git a/block/vpc.c b/block/vpc.c
>>> index 416f489..35ac3fd 100644
>>> --- a/block/vpc.c
>>> +++ b/block/vpc.c
>>> @@ -585,7 +585,7 @@ static int vpc_create(const char *filename, 
>> QEMUOptionParameter *options)
>>>
>>>      memcpy(dyndisk_header->magic, "cxsparse", 8);
>>>
>>> -    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
>>> +    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
>>>      dyndisk_header->table_offset = be64_to_cpu(3 * 512);
>>>      dyndisk_header->version = be32_to_cpu(0x00010000);
>>>      dyndisk_header->block_size = be32_to_cpu(block_size);
>>
>> Can you please add a short comment explaining why we deviate from the
>> specification in this point? If someone notices the discrepancy later,
>> he shouldn't have to dig up the git commit message.
>>
>> Kevin
>  
> Using qemu-img to convert a raw image to a vhd image seems to work fine accept that only 4 bytes of the 8 byte Dynamic Disk Header Data Offset field are initialized.  This seems correct according to the specification but tools like Citrix's vhd-util fail when querying the image due to the shortened initialization.  Upon further analysis, it was determined that Microsoft also initializes all 8 bytes when generating a vhd image on windows.  Initializing all 8 bytes seems more logical and may indicate that Microsoft needs to update the specification because the requirement for this field changed or that it is simply an oversight that has never been noticed.

Right, I'm not arguing that your change is wrong. What I'm asking for is
just something like:

-    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
+    /* Note: The spec is actually wrong here, it says 0xFFFFFFFF, but
MS tools expect all 64 bits to be set */
+    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);

Kevin

Am 09.11.2011 16:42, schrieb Charles Arnold:
>>>> On 11/9/2011 at 02:46 AM, in message <4EBA4BEC.5040401@redhat.com>, Kevin Wolf
> <kwolf@redhat.com> wrote: 
>> Am 08.11.2011 20:25, schrieb Charles Arnold:
>>> The Data Offset field in the Dynamic Disk Header is an 8 byte field.
>>> Although the specification (2006-10-11) gives an example of initializing
>>> only the first 4 bytes, images generated by Microsoft on Windows initialize 
>>> all 8 bytes.
>>>
>>> Failure to initialize all 8 bytes results in errors from utilities
>>> that check specifically for the complete Data Offset field initialization.
>>>
>>> Signed-off-by: Charles Arnold <carnold@suse.com>
>>>
>>> diff --git a/block/vpc.c b/block/vpc.c
>>> index 416f489..35ac3fd 100644
>>> --- a/block/vpc.c
>>> +++ b/block/vpc.c
>>> @@ -585,7 +585,7 @@ static int vpc_create(const char *filename, 
>> QEMUOptionParameter *options)
>>>
>>>      memcpy(dyndisk_header->magic, "cxsparse", 8);
>>>
>>> -    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFF);
>>> +    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
>>>      dyndisk_header->table_offset = be64_to_cpu(3 * 512);
>>>      dyndisk_header->version = be32_to_cpu(0x00010000);
>>>      dyndisk_header->block_size = be32_to_cpu(block_size);
>>
>> Can you please add a short comment explaining why we deviate from the
>> specification in this point? If someone notices the discrepancy later,
>> he shouldn't have to dig up the git commit message.
>>
>> Kevin
>  
> Using qemu-img to convert a raw image to a vhd image seems to work fine accept that only 4 bytes of the 8 byte Dynamic Disk Header Data Offset field are initialized.  This seems correct according to the specification but tools like Citrix's vhd-util fail when querying the image due to the shortened initialization.  Upon further analysis, it was determined that Microsoft also initializes all 8 bytes when generating a vhd image on windows.  Initializing all 8 bytes seems more logical and may indicate that Microsoft needs to update the specification because the requirement for this field changed or that it is simply an oversight that has never been noticed.

Charles, I think Kevin meant inserting a C comment above that line. :)

Maybe: /* Microsoft and Citrix tools expect data_offset to be eight
bytes 0xff although this does not match the spec. */

Regards,
Andreas