Re: [Qemu-devel] [PATCH v3 14/29] block: Switch BdrvTrackedRequest to byte granularity

["Benoît Canet" <benoit.canet@irqsave.net>]

Le Friday 17 Jan 2014 à 15:15:04 (+0100), Kevin Wolf a écrit :
> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> ---
>  block.c                   | 52 +++++++++++++++++++++++++++++++----------------
>  block/backup.c            |  7 ++++++-
>  include/block/block_int.h |  4 ++--
>  3 files changed, 42 insertions(+), 21 deletions(-)
> 
> diff --git a/block.c b/block.c
> index 328f592..85f28ab 100644
> --- a/block.c
> +++ b/block.c
> @@ -2037,13 +2037,13 @@ static void tracked_request_end(BdrvTrackedRequest *req)
>   */
>  static void tracked_request_begin(BdrvTrackedRequest *req,
>                                    BlockDriverState *bs,
> -                                  int64_t sector_num,
> -                                  int nb_sectors, bool is_write)
> +                                  int64_t offset,
> +                                  unsigned int bytes, bool is_write)
>  {
>      *req = (BdrvTrackedRequest){
>          .bs = bs,
> -        .sector_num = sector_num,
> -        .nb_sectors = nb_sectors,
> +        .offset = offset,
> +        .bytes = bytes,
>          .is_write = is_write,
>          .co = qemu_coroutine_self(),
>      };
> @@ -2074,25 +2074,43 @@ void bdrv_round_to_clusters(BlockDriverState *bs,
>      }
>  }
>  
> +static void round_bytes_to_clusters(BlockDriverState *bs,
> +                                    int64_t offset, unsigned int bytes,
> +                                    int64_t *cluster_offset,
> +                                    unsigned int *cluster_bytes)
> +{
> +    BlockDriverInfo bdi;
> +
> +    if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) {
> +        *cluster_offset = offset;
> +        *cluster_bytes = bytes;
> +    } else {
> +        *cluster_offset = QEMU_ALIGN_DOWN(offset, bdi.cluster_size);
> +        *cluster_bytes = QEMU_ALIGN_UP(offset - *cluster_offset + bytes,
> +                                       bdi.cluster_size);
> +    }
> +}
> +
>  static bool tracked_request_overlaps(BdrvTrackedRequest *req,
> -                                     int64_t sector_num, int nb_sectors) {
> +                                     int64_t offset, unsigned int bytes)
> +{
>      /*        aaaa   bbbb */
> -    if (sector_num >= req->sector_num + req->nb_sectors) {
> +    if (offset >= req->offset + req->bytes) {
>          return false;
>      }
>      /* bbbb   aaaa        */
> -    if (req->sector_num >= sector_num + nb_sectors) {
> +    if (req->offset >= offset + bytes) {
>          return false;
>      }
>      return true;
>  }
>  
>  static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs,
> -        int64_t sector_num, int nb_sectors)
> +        int64_t offset, unsigned int bytes)
>  {
>      BdrvTrackedRequest *req;
> -    int64_t cluster_sector_num;
> -    int cluster_nb_sectors;
> +    int64_t cluster_offset;
> +    unsigned int cluster_bytes;
>      bool retry;
>  
>      /* If we touch the same cluster it counts as an overlap.  This guarantees
> @@ -2101,14 +2119,12 @@ static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs,
>       * CoR read and write operations are atomic and guest writes cannot
>       * interleave between them.
>       */
> -    bdrv_round_to_clusters(bs, sector_num, nb_sectors,
> -                           &cluster_sector_num, &cluster_nb_sectors);
> +    round_bytes_to_clusters(bs, offset, bytes, &cluster_offset, &cluster_bytes);
>  
>      do {
>          retry = false;
>          QLIST_FOREACH(req, &bs->tracked_requests, list) {
> -            if (tracked_request_overlaps(req, cluster_sector_num,
> -                                         cluster_nb_sectors)) {
> +            if (tracked_request_overlaps(req, cluster_offset, cluster_bytes)) {
>                  /* Hitting this means there was a reentrant request, for
>                   * example, a block driver issuing nested requests.  This must
>                   * never happen since it means deadlock.
> @@ -2728,10 +2744,10 @@ static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs,
>      }
>  
>      if (bs->copy_on_read_in_flight) {
> -        wait_for_overlapping_requests(bs, sector_num, nb_sectors);
> +        wait_for_overlapping_requests(bs, offset, bytes);
>      }
>  
> -    tracked_request_begin(&req, bs, sector_num, nb_sectors, false);
> +    tracked_request_begin(&req, bs, offset, bytes, false);
>  
>      if (flags & BDRV_REQ_COPY_ON_READ) {
>          int pnum;
> @@ -2980,10 +2996,10 @@ static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs,
>      assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
>  
>      if (bs->copy_on_read_in_flight) {
> -        wait_for_overlapping_requests(bs, sector_num, nb_sectors);
> +        wait_for_overlapping_requests(bs, offset, bytes);
>      }
>  
> -    tracked_request_begin(&req, bs, sector_num, nb_sectors, true);
> +    tracked_request_begin(&req, bs, offset, bytes, true);
>  
>      ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req);
>  
> diff --git a/block/backup.c b/block/backup.c
> index 0198514..15a2e55 100644
> --- a/block/backup.c
> +++ b/block/backup.c
> @@ -181,8 +181,13 @@ static int coroutine_fn backup_before_write_notify(
>          void *opaque)
>  {
>      BdrvTrackedRequest *req = opaque;
> +    int64_t sector_num = req->offset >> BDRV_SECTOR_BITS;
> +    int nb_sectors = req->bytes >> BDRV_SECTOR_BITS;
>  
> -    return backup_do_cow(req->bs, req->sector_num, req->nb_sectors, NULL);
> +    assert((req->offset & (BDRV_SECTOR_SIZE - 1)) == 0);
> +    assert((req->bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
> +
> +    return backup_do_cow(req->bs, sector_num, nb_sectors, NULL);
>  }
>  
>  static void backup_set_speed(BlockJob *job, int64_t speed, Error **errp)
> diff --git a/include/block/block_int.h b/include/block/block_int.h
> index fb1d355..97a4d23 100644
> --- a/include/block/block_int.h
> +++ b/include/block/block_int.h
> @@ -57,8 +57,8 @@
>  
>  typedef struct BdrvTrackedRequest {
>      BlockDriverState *bs;
> -    int64_t sector_num;
> -    int nb_sectors;
> +    int64_t offset;
> +    unsigned int bytes;
>      bool is_write;
>      QLIST_ENTRY(BdrvTrackedRequest) list;
>      Coroutine *co; /* owner, used for deadlock detection */
> -- 
> 1.8.1.4
> 
> 
Reviewed-by: Benoit Canet <benoit@irqsave.net>