Re: [Qemu-devel] [PATCH v5 3/3] qcow2: add zstd cluster compression

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1.1: Type: text/plain, Size: 6691 bytes --]

On 9/4/19 10:29 AM, Denis Plotnikov wrote:
> zstd significantly reduces cluster compression time.
> It provides better compression performance maintaining
> the same level of compression ratio in comparison with
> zlib, which, at the moment, has been the only compression
> method available.
> 
> The performance test results:
> Test compresses and decompresses qemu qcow2 image with just
> installed rhel-7.6 guest.
> Image cluster size: 64K. Image on disk size: 2.2G
> 

> 
> Signed-off-by: Denis Plotnikov <dplotnikov@virtuozzo.com>
> ---

> +static ssize_t qcow2_zstd_compress(void *dest, size_t dest_size,
> +                                   const void *src, size_t src_size)
> +{
> +    ssize_t ret;
> +    uint32_t *c_size = dest;

Potentially unaligned pointer...

> +    /* steal some bytes to store compressed chunk size */
> +    char *d_buf = ((char *) dest) + sizeof(*c_size);
> +
> +    /* sanity check that we can store the compressed data length */
> +    if (dest_size < sizeof(*c_size)) {
> +        return -ENOMEM;
> +    }
> +
> +    dest_size -= sizeof(*c_size);
> +
> +    ret = ZSTD_compress(d_buf, dest_size, src, src_size, 5);
> +
> +    if (ZSTD_isError(ret)) {
> +        if (ZSTD_getErrorCode(ret) == ZSTD_error_dstSize_tooSmall) {
> +            return -ENOMEM;
> +        } else {
> +            return -EIO;
> +        }
> +    }
> +
> +    /* store the compressed chunk size in the very beginning of the buffer */
> +    *c_size = cpu_to_be32(ret);

...and you are storing into it.  You are using the wrong conversion
function; you want stl_be_p(dest, ret) or similar.

> +static ssize_t qcow2_zstd_decompress(void *dest, size_t dest_size,
> +                                     const void *src, size_t src_size)
> +{
> +    ssize_t ret;
> +    /*
> +     * zstd decompress wants to know the exact length of the data
> +     * for that purpose, on the compression the length is stored in

s/data for/data. For/
s/on the compression/on compression,/

> +     * the very beginning of the compressed buffer
> +     */
> +    uint32_t s_size;
> +    const char *s_buf = ((const char *) src) + sizeof(s_size);
> +
> +    /* sanity check that we can read the content length */
> +    if (src_size < sizeof(s_size)) {

Should this use <=?  After all, I seriously doubt you can get a 0-byte
compression stream.

> +        return -EIO;
> +    }
> +
> +    s_size = be32_to_cpu(*(const uint32_t *) src);

As written, this looks like you may be dereferencing an unaligned
pointer.  It so happens that be32_to_cpu() applies & to your * to get
back at the raw pointer, and then is careful to handle unaligned
pointers, so it works; but it would look a lot nicer as merely:

s_size = be32_to_cpu(src);

> +
> +    /* sanity check that the buffer is big enough to read the content */
> +    if (src_size - sizeof(s_size) < s_size) {

Why < and not !=?  As written, you are silently ignoring trailing
garbage, instead of treating it as a client that did not write the data
correctly.

> +++ b/docs/interop/qcow2.txt
> @@ -181,6 +181,7 @@ in the description of a field.
>                      must be set.
>                      Available compression type values:
>                          0: zlib <https://www.zlib.net/> (default)
> +                        1: zstd <http://github.com/facebook/zstd>

At this point, this listing is almost redundant with the more-detailed
header below.  Maybe it is worth just forward referencing that section
for a listing of valid values, and then mentioning the values 0 and 1 in
that section?

>  
>  Directly after the image header, optional sections called header extensions can
>  be stored. Each extension has a structure like the following:
> @@ -536,6 +537,9 @@ Compressed Clusters Descriptor (x = 62 - (cluster_bits - 8)):
>                      Another compressed cluster may map to the tail of the final
>                      sector used by this compressed cluster.
>  
> +                    The layout of the compressed data depends on the compression
> +                    type used for the image (see compressed cluster layout).
> +
>  If a cluster is unallocated, read requests shall read the data from the backing
>  file (except if bit 0 in the Standard Cluster Descriptor is set). If there is
>  no backing file or the backing file is smaller than the image, they shall read
> @@ -788,3 +792,19 @@ In the image file the 'enabled' state is reflected by the 'auto' flag. If this
>  flag is set, the software must consider the bitmap as 'enabled' and start
>  tracking virtual disk changes to this bitmap from the first write to the
>  virtual disk. If this flag is not set then the bitmap is disabled.
> +
> +=== Compressed cluster layout ===

As written, you have made this a child to '== Bitmaps ==' (and sibling
to '=== Dirty tracking bitmaps ==='); that feels wrong.  I would place
this subsection belong under '== Cluster mapping ==' right after the
'Compressed Clusters Descriptor'.

> +
> +The compressed cluster data may have a different layout depending on the
> +compression type used for the image, and store specific data for the particular
> +compression type.
> +
> +Compressed data layout for the available compression types:

Wordy; maybe:

The compressed cluster data has a layout depending on the compression
type used for the image, as follows:

> +(x = data_space_length - 1)
> +
> +    zlib <http://zlib.net/>:
> +        Byte  0 -  x:     the compressed data content
> +                          all the space provided used for compressed data

Worth a mention that this is compression type 0?

> +    zstd <http://github.com/facebook/zstd>:
> +        Byte  0 -  3:     the length of compressed data in bytes
> +              4 -  x:     the compressed data content

Worth a mention that this is compression type 1?

> diff --git a/qapi/block-core.json b/qapi/block-core.json
> index 2c002ca6a9..9e458d5b40 100644
> --- a/qapi/block-core.json
> +++ b/qapi/block-core.json
> @@ -4283,11 +4283,12 @@
>  # Compression type used in qcow2 image file
>  #
>  # @zlib:  zlib compression, see <http://zlib.net/>
> +# @zstd:  zstd compression, see <http://github.com/facebook/zstd>
>  #
>  # Since: 4.2
>  ##
>  { 'enum': 'Qcow2CompressionType',
> -  'data': [ 'zlib' ] }
> +  'data': [ 'zlib', { 'name': 'zstd', 'if': 'defined(CONFIG_ZSTD)' } ] }
>  
>  ##
>  # @BlockdevCreateOptionsQcow2:
> 

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]