[sheepdog] [Qemu-devel] [PATCH v2 2/2] sheepdog: support user-defined redundancy option

[Benoît Canet]

Le Tuesday 29 Oct 2013 à 16:25:52 (+0800), Liu Yuan a écrit :
> Sheepdog support two kinds of redundancy, full replication and erasure coding.
> 
> # create a fully replicated vdi with x copies
>  -o redundancy=x (1 <= x <= SD_MAX_COPIES)
> 
> # create a erasure coded vdi with x data strips and y parity strips
>  -o redundancy=x:y (x must be one of {2,4,8,16} and 1 <= y < SD_EC_MAX_STRIP)
> 
> E.g, to convert a vdi into sheepdog vdi 'test' with 8:3 erasure coding scheme
> 
> $ qemu-img convert -o redundancy=8:3 linux-0.2.img sheepdog:test
> 
> Cc: Kevin Wolf <kwolf at redhat.com>
> Cc: Stefan Hajnoczi <stefanha at redhat.com>
> Signed-off-by: Liu Yuan <namei.unix at gmail.com>
> ---
>  block/sheepdog.c          |   78 ++++++++++++++++++++++++++++++++++++++++++++-
>  include/block/block_int.h |    1 +
>  2 files changed, 78 insertions(+), 1 deletion(-)
> 
> diff --git a/block/sheepdog.c b/block/sheepdog.c
> index e66d2f8..bd7cfd6 100644
> --- a/block/sheepdog.c
> +++ b/block/sheepdog.c
> @@ -91,6 +91,14 @@
>  #define SD_NR_VDIS   (1U << 24)
>  #define SD_DATA_OBJ_SIZE (UINT64_C(1) << 22)
>  #define SD_MAX_VDI_SIZE (SD_DATA_OBJ_SIZE * MAX_DATA_OBJS)
> +/*
> + * For erasure coding, we use at most SD_EC_MAX_STRIP for data strips and
> + * (SD_EC_MAX_STRIP - 1) for parity strips
> + *
> + * SD_MAX_COPIES is sum of number of dats trips and parity strips.
> + */
> +#define SD_EC_MAX_STRIP 16
> +#define SD_MAX_COPIES (SD_EC_MAX_STRIP * 2 - 1)
>  
>  #define SD_INODE_SIZE (sizeof(SheepdogInode))
>  #define CURRENT_VDI_ID 0
> @@ -1446,6 +1454,65 @@ out:
>      return ret;
>  }
>  
> +static int64_t is_numeric(const char *s)
> +{
> +    char *end;
> +    return strtosz_suffix(s, &end, STRTOSZ_DEFSUFFIX_B);
> +}
> +
> +/*
> + * Sheepdog support two kinds of redundancy, full replication and erasure
> + * coding.
> + *
> + * # create a fully replicated vdi with x copies
> + * -o redundancy=x (1 <= x <= SD_MAX_COPIES)
> + *
> + * # create a erasure coded vdi with x data strips and y parity strips
> + * -o redundancy=x:y (x must be one of {2,4,8,16} and 1 <= y < SD_EC_MAX_STRIP)
> + */
> +static int parse_redundancy(BDRVSheepdogState *s, const char *opt)
> +{
> +    struct SheepdogInode *inode = &s->inode;
> +    const char *n1, *n2;
> +    uint8_t copy, parity;
> +    char p[10];
> +
> +    strncpy(p, opt, sizeof(p));
> +    n1 = strtok(p, ":");
> +    n2 = strtok(NULL, ":");
> +
> +    if ((n1 && !is_numeric(n1)) || (n2 && !is_numeric(n2))) {
> +        return -EINVAL;
> +    }
> +
> +    copy = strtol(n1, NULL, 10);
> +    if (copy > SD_MAX_COPIES) {
> +        return -EINVAL;
> +    }
> +    if (!n2) {
> +        inode->copy_policy = 0;
> +        inode->nr_copies = copy;
> +    }
> +
> +    if (copy != 2 && copy != 4 && copy != 8 && copy != 16) {
> +        return -EINVAL;
> +    }
> +
> +    parity = strtol(n2, NULL, 10);
> +    if (parity >= SD_EC_MAX_STRIP || parity == 0) {
> +        return -EINVAL;
> +    }
> +
> +    /*
> +     * 4 bits for parity and 4 bits for data.
> +     * We have to compress upper data bits because it can't represent 16
> +     */
> +    inode->copy_policy = ((copy / 2) << 4) + parity;
> +    inode->nr_copies = copy + parity;
> +
> +    return 0;
> +}
> +
>  static int sd_create(const char *filename, QEMUOptionParameter *options,
>                       Error **errp)
>  {
> @@ -1486,6 +1553,11 @@ static int sd_create(const char *filename, QEMUOptionParameter *options,
>                  ret = -EINVAL;
>                  goto out;
>              }
> +        } else if (!strcmp(options->name, BLOCK_OPT_REDUNDANCY)) {
> +            ret = parse_redundancy(s, options->value.s);
> +            if (ret < 0) {
> +                goto out;
> +            }
>          }
>          options++;
>      }
> @@ -1528,7 +1600,6 @@ static int sd_create(const char *filename, QEMUOptionParameter *options,
>          bdrv_unref(bs);
>      }
>  
> -    /* TODO: allow users to specify copy number */
>      ret = do_sd_create(s, &vid, 0);
>      if (!prealloc || ret) {
>          goto out;
> @@ -2332,6 +2403,11 @@ static QEMUOptionParameter sd_create_options[] = {
>          .type = OPT_STRING,
>          .help = "Preallocation mode (allowed values: off, full)"
>      },
> +    {
> +        .name = BLOCK_OPT_REDUNDANCY,
> +        .type = OPT_STRING,
> +        .help = "Redundancy of the image"
> +    },
>      { NULL }
>  };
>  
> diff --git a/include/block/block_int.h b/include/block/block_int.h
> index a48731d..b90862f 100644
> --- a/include/block/block_int.h
> +++ b/include/block/block_int.h
> @@ -53,6 +53,7 @@
>  #define BLOCK_OPT_COMPAT_LEVEL      "compat"
>  #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
>  #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
> +#define BLOCK_OPT_REDUNDANCY        "redundancy"
>  
>  typedef struct BdrvTrackedRequest {
>      BlockDriverState *bs;
> -- 
> 1.7.9.5
> 
> 

Hi,

Perhaps it would be better to implement theses options with the new block driver
options mechanism that Kevin implemented.

>From a user perspective it look like this.
  -drive if=virtio,file.driver=quorum,\
    file.children.0.file.filename=1.qcow2,\
    file.children.1.file.filename=2.qcow2,\
    file.children.2.file.filename=3.qcow2,\
    file.vote_threshold=3

I don't know if it would work with protocols but that would uniformize the
options passing style accross multiple block drivers.

I have an under review patch on the list using this mechanism, see:
"[PATCH V9 11/11] quorum: Add quorum_open() and quorum_close()."

Best regards

Benoît