Re: [Qemu-devel] [PATCH v7] spec: add qcow2 bitmaps extension specification

[John Snow <jsnow@redhat.com>]

On 01/14/2016 05:08 PM, Eric Blake wrote:
> On 01/11/2016 06:05 AM, Vladimir Sementsov-Ogievskiy wrote:
>> The new feature for qcow2: storing bitmaps.
>>
>> This patch adds new header extension to qcow2 - Bitmaps Extension. It
>> provides an ability to store virtual disk related bitmaps in a qcow2
>> image. For now there is only one type of such bitmaps: Dirty Tracking
>> Bitmap, which just tracks virtual disk changes from some moment.
>>
>> Note: Only bitmaps, relative to the virtual disk, stored in qcow2 file,
>> should be stored in this qcow2 file. The size of each bitmap
>> (considering its granularity) is equal to virtual disk size.
>>
>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>> ---
>>
> 
>> @@ -166,6 +178,34 @@ the header extension data. Each entry look like this:
>>                      terminated if it has full length)
>>  
>>  
>> +== Bitmaps extension ==
> 
>> +          0 -  3:  nb_bitmaps
>> +                   The number of bitmaps contained in the image. Must be
>> +                   greater than or equal to 1.
>> +
>> +                   Note: Qemu currently only supports up to 65535 bitmaps per
>> +                   image.
>> +
>> +          4 -  7:  bitmap_directory_size
>> +                   Size of the bitmap directory in bytes. It is the cumulative
>> +                   size of all (nb_bitmaps) bitmap headers.
> 
> Only 4 bytes - if we ever raise our 64k entry restriction (nb_bitmaps),
> could we run into an image that has so many directory entries as to make
> the directory itself spill past 4G?  But I don't think it is likely, so
> I can live with your choice.
> 

"We'll never need this!"

I hope someone in 2082 is reading this right now and is quite angry.

(But really, I can't foresee needing this many per each drive -- and if
we do, we have external storage mechanisms in development to handle such
wild cases.)

>> +
>> +== Bitmaps ==
>> +
>> +As mentioned above, the bitmaps extension provides the ability to store bitmaps
>> +related a virtual disk. This section describes how these bitmaps are stored.
>> +
>> +Note: all bitmaps are related to the virtual disk stored in this image.
>> +
>> +=== Bitmap directory ===
>> +
>> +Each bitmap saved in the image is described in a bitmap directory entry. The
>> +bitmap directory is a contiguous area in the image file, whose starting offset
>> +and length are given by the header extension fields bitmap_directory_offset and
>> +bitmap_directory_size. The entries of the bitmap directory have variable
>> +length, depending on the length of the bitmap name and extra data. These
>> +entries are also called bitmap headers.
>> +
>> +Structure of a bitmap directory entry:
>> +
>> +    Byte 0 -  7:    bitmap_table_offset
>> +                    Offset into the image file at which the bitmap table
>> +                    (described below) for the bitmap starts. Must be aligned to
>> +                    a cluster boundary.
>> +
>> +         8 - 11:    bitmap_table_size
>> +                    Number of entries in the bitmap table of the bitmap.
> 
> Should this be the size in bytes, instead of the number of entries? But

For what benefit? We can calculate either from the other, and this gives
us a better resolution.

> at least the entries are fixed width of 8 bytes each, so this lets you
> get a bitmap table up to 32G bytes rather than just 4G in size.  (Let's
> see here - if we have 32G bytes in the bitmap table, that means 4G
> clusters occupied by the bitmap itself; in the worst case of 512-byte
> clusters and granularity 0, that is a maximum bitmap size of 2T
> describing 16T of virtual guest image; with larger cluster size and/or
> larger granularity, we cover a lot more virtual guest space with less
> bitmap size; so I guess we aren't too worried about running out of space?).
> 

Yes, worst case of g=0 and cluster size of 512 bytes, we can get 2T
bitmaps describing 16T of virtual data.

"default case" of 64K clusters and 64K granularity: 256TiB bitmaps
describing ... let's see ... if my math is right, 128EiB?

We're probably fine :)

(Cue future space-person from 2159 wondering how I could have ever been
so naive. Sorry, future space-person!)

>> +        20 - 23:    extra_data_size
>> +                    Size of type-specific extra data.
>> +
>> +                    For now, as no extra data is defined, extra_data_size is
>> +                    reserved and must be zero.
>> +
>> +        variable:   Type-specific extra data for the bitmap.
> 
> I'd write this as:
>            variable:   extra_data
>                        Type-specific extra data for the bitmap,
>                        occupying extra_data_size bytes.
> 
>> +
>> +        variable:   The name of the bitmap (not null terminated). Must be
>> +                    unique among all bitmap names within the bitmaps extension.
>> +
>> +        variable:   Padding to round up the bitmap directory entry size to the
>> +                    next multiple of 8.
> 
> Should we require the padding to be all NUL bytes?  (We aren't
> consistent on whether we require that for other locations of padding in
> the spec, so that could be a followup patch).
> 
>> +
>> +=== Bitmap table ===
>> +
>> +Bitmaps are stored using a one-level structure (as opposed to two-level
>> +structure like for refcounts and guest clusters mapping) for the mapping of
>> +bitmap data to host clusters. This structure is called the bitmap table.
> 
> Possible wording tweak:
> Bitmaps are stored using a one-level structure (as opposed to the
> two-level structures for refcounts and guest cluster mapping), and are
> used for the mapping of bitmap data to host clusters
> 
>> +
>> +Each bitmap table has a variable size (stored in the bitmap directory Entry)
> 
> Does 'Entry' still need to be capitalized?
> 
>> +and may use multiple clusters, however, it must be contiguous in the image
>> +file.
>> +
>> +Structure of a bitmap table entry:
>> +
>> +    Bit       0:    Reserved and must be zero if bits 9 - 55 are non-zero.
>> +                    If bits 9 - 55 are zero:
>> +                      0: Cluster should be read as all zeros.
>> +                      1: Cluster should be read as all ones.
>> +
>> +         1 -  8:    Reserved and must be zero.
>> +
>> +         9 - 55:    Bits 9 - 55 of the host cluster offset. Must be aligned to
>> +                    a cluster boundary. If the offset is 0, the cluster is
>> +                    unallocated; in that case, bit 0 determines how this
>> +                    cluster should be treated when read from.
> 
> Possible wording tweak:
> s/when read from/during reads/.
> 
>> +
>> +        56 - 63:    Reserved and must be zero.
>> +
>> +=== Bitmap data ===
>> +
>> +As noted above, bitmap data is stored in separate clusters, described by the
>> +bitmap table. Given an offset (in bytes) into the bitmap data, the offset into
>> +the image file can be obtained as follows:
>> +
>> +    image_offset =
>> +        bitmap_table[bitmap_data_offset / cluster_size] +
>> +            (bitmap_data_offset % cluster_size)
>> +
>> +This offset is not defined if bits 9 - 55 of bitmap table entry are zero (see
>> +above).
>> +
>> +Given an offset byte_nr into the virtual disk and the bitmap's granularity, the
>> +bit offset into the bitmap can be calculated like this:
>> +
>> +    bit_offset =
>> +        image_offset(byte_nr / granularity / 8) * 8 +
>> +            (byte_nr / granularity) % 8
>> +
>> +If the size of the bitmap data is not a multiply of cluster size then the last
> 
> s/multiply of cluster size/multiple of the cluster size,/
> 
>> +cluster of the bitmap data contains some unused tail bits. These bits must be
>> +zero.
>>
> 

Thanks!

--js