Re: [Qemu-devel] [PATCH 01/22] block: Move initialisation of BlockLimits to bdrv_refresh_limits()

[Peter Lieven]

Am 11.12.2013 22:08, schrieb Kevin Wolf:
> This function separates filling the BlockLimits from bdrv_open(), which
> allows it to call it from other operations which may change the limits
> (e.g. modifications to the backing file chain or bdrv_reopen)
>
> Signed-off-by: Kevin Wolf <address@hidden>
> ---
>  block.c                   | 21 ++++++++++++
>  block/iscsi.c             | 87 
> ++++++++++++++++++++++++++++-------------------
>  block/qcow2.c             | 11 +++++-
>  block/qed.c               | 11 +++++-
>  block/vmdk.c              | 22 +++++++++---
>  include/block/block_int.h |  2 ++
>  6 files changed, 113 insertions(+), 41 deletions(-)
>
> diff --git a/block.c b/block.c
> index 13f001a..c32d856 100644
> --- a/block.c
> +++ b/block.c
> @@ -479,6 +479,21 @@ int bdrv_create_file(const char* filename, 
> QEMUOptionParameter *options,
>      return ret;
>  }
>  
> +static int bdrv_refresh_limits(BlockDriverState *bs)
> +{
> +    BlockDriver *drv = bs->drv;
> +
> +    memset(&bs->bl, 0, sizeof(bs->bl));
> +
> +    if (!drv) {
> +        return 0;
> +    } else if (drv->bdrv_refresh_limits) {
> +        return drv->bdrv_refresh_limits(bs);
> +    }
> +
> +    return 0;
> +}
> +
>  /*
>   * Create a uniquely-named empty temporary file.
>   * Return 0 upon success, otherwise a negative errno value.
> @@ -833,6 +848,8 @@ static int bdrv_open_common(BlockDriverState *bs, 
> BlockDriverState *file,
>          goto free_and_fail;
>      }
>  
> +    bdrv_refresh_limits(bs);
> +
>  #ifndef _WIN32
>      if (bs->is_temporary) {
>          assert(bs->filename[0] != '\0');
> @@ -1018,6 +1035,10 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict 
> *options, Error **errp)
>      }
>      pstrcpy(bs->backing_file, sizeof(bs->backing_file),
>              bs->backing_hd->file->filename);
> +
> +    /* Recalculate the BlockLimits with the backing file */
> +    bdrv_refresh_limits(bs);
> +
>      return 0;
>  }
>  
> diff --git a/block/iscsi.c b/block/iscsi.c
> index 829d444..f3ded8c 100644
> --- a/block/iscsi.c
> +++ b/block/iscsi.c
> @@ -1425,6 +1425,56 @@ static int iscsi_open(BlockDriverState *bs, QDict 
> *options, int flags,
>          task = NULL;
>      }
>  
> +#if defined(LIBISCSI_FEATURE_NOP_COUNTER)
> +    /* Set up a timer for sending out iSCSI NOPs */
> +    iscsilun->nop_timer = timer_new_ms(QEMU_CLOCK_REALTIME, 
> iscsi_nop_timed_event, iscsilun);
> +    timer_mod(iscsilun->nop_timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + 
> NOP_INTERVAL);
> +#endif
> +
> +out:
> +    qemu_opts_del(opts);
> +    if (initiator_name != NULL) {
> +        g_free(initiator_name);
> +    }
> +    if (iscsi_url != NULL) {
> +        iscsi_destroy_url(iscsi_url);
> +    }
> +    if (task != NULL) {
> +        scsi_free_scsi_task(task);
> +    }
> +
> +    if (ret) {
> +        if (iscsi != NULL) {
> +            iscsi_destroy_context(iscsi);
> +        }
> +        memset(iscsilun, 0, sizeof(IscsiLun));
> +    }
> +    return ret;
> +}
> +
> +static void iscsi_close(BlockDriverState *bs)
> +{
> +    IscsiLun *iscsilun = bs->opaque;
> +    struct iscsi_context *iscsi = iscsilun->iscsi;
> +
> +    if (iscsilun->nop_timer) {
> +        timer_del(iscsilun->nop_timer);
> +        timer_free(iscsilun->nop_timer);
> +    }
> +    qemu_aio_set_fd_handler(iscsi_get_fd(iscsi), NULL, NULL, NULL);
> +    iscsi_destroy_context(iscsi);
> +    g_free(iscsilun->zeroblock);
> +    memset(iscsilun, 0, sizeof(IscsiLun));
> +}
> +
> +static int iscsi_refresh_limits(BlockDriverState *bs)
> +{
> +    IscsiLun *iscsilun = bs->opaque;
> +    struct scsi_task *task = NULL;
> +    int ret;
> +
> +    memset(&bs->bl, 0, sizeof(bs->bl));
> +
you do that memset in bdrv_refresh_limits already.
>      if (iscsilun->lbp.lbpu || iscsilun->lbp.lbpws) {
>          struct scsi_inquiry_block_limits *inq_bl;
>          task = iscsi_do_inquiry(iscsilun->iscsi, iscsilun->lun, 1,
> @@ -1462,48 +1512,14 @@ static int iscsi_open(BlockDriverState *bs, QDict 
> *options, int flags,

iscsi_do_inquiry does a sync call to request the pages from the target.
you can only do that here if you can guarantee that iscsi_refresh_limits will
be called with no pending requests in-flight otherwise this will end in a mess.

otherwise i would propose to leave the inquiry call ins iscsi_open and just fill
the bs->bl struc tin iscsi_refresh_limits. the problem is if we make this async
we have to handle all possible I/O callbacks whiile in iscsi_refresh_limits.
i do not see how an iscsi target can change its limits while a target is 
mounted.

Peter

>                                                       iscsilun);
>      }
>  
> -#if defined(LIBISCSI_FEATURE_NOP_COUNTER)
> -    /* Set up a timer for sending out iSCSI NOPs */
> -    iscsilun->nop_timer = timer_new_ms(QEMU_CLOCK_REALTIME, 
> iscsi_nop_timed_event, iscsilun);
> -    timer_mod(iscsilun->nop_timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + 
> NOP_INTERVAL);
> -#endif
> -
> +    ret = 0;
>  out:
> -    qemu_opts_del(opts);
> -    if (initiator_name != NULL) {
> -        g_free(initiator_name);
> -    }
> -    if (iscsi_url != NULL) {
> -        iscsi_destroy_url(iscsi_url);
> -    }
>      if (task != NULL) {
>          scsi_free_scsi_task(task);
>      }
> -
> -    if (ret) {
> -        if (iscsi != NULL) {
> -            iscsi_destroy_context(iscsi);
> -        }
> -        memset(iscsilun, 0, sizeof(IscsiLun));
> -    }
>      return ret;
>  }
>  
> -static void iscsi_close(BlockDriverState *bs)
> -{
> -    IscsiLun *iscsilun = bs->opaque;
> -    struct iscsi_context *iscsi = iscsilun->iscsi;
> -
> -    if (iscsilun->nop_timer) {
> -        timer_del(iscsilun->nop_timer);
> -        timer_free(iscsilun->nop_timer);
> -    }
> -    qemu_aio_set_fd_handler(iscsi_get_fd(iscsi), NULL, NULL, NULL);
> -    iscsi_destroy_context(iscsi);
> -    g_free(iscsilun->zeroblock);
> -    memset(iscsilun, 0, sizeof(IscsiLun));
> -}
> -
>  static int iscsi_truncate(BlockDriverState *bs, int64_t offset)
>  {
>      IscsiLun *iscsilun = bs->opaque;
> @@ -1616,6 +1632,7 @@ static BlockDriver bdrv_iscsi = {
>      .bdrv_getlength  = iscsi_getlength,
>      .bdrv_get_info   = iscsi_get_info,
>      .bdrv_truncate   = iscsi_truncate,
> +    .bdrv_refresh_limits = iscsi_refresh_limits,
>  
>  #if defined(LIBISCSI_FEATURE_IOVECTOR)
>      .bdrv_co_get_block_status = iscsi_co_get_block_status,
> diff --git a/block/qcow2.c b/block/qcow2.c
> index f29aa88..f40355c 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -718,7 +718,6 @@ static int qcow2_open(BlockDriverState *bs, QDict 
> *options, int flags,
>      }
>  
>      qemu_opts_del(opts);
> -    bs->bl.write_zeroes_alignment = s->cluster_sectors;
>  
>      if (s->use_lazy_refcounts && s->qcow_version < 3) {
>          error_setg(errp, "Lazy refcounts require a qcow2 image with at least 
> "
> @@ -751,6 +750,15 @@ static int qcow2_open(BlockDriverState *bs, QDict 
> *options, int flags,
>      return ret;
>  }
>  
> +static int qcow2_refresh_limits(BlockDriverState *bs)
> +{
> +    BDRVQcowState *s = bs->opaque;
> +
> +    bs->bl.write_zeroes_alignment = s->cluster_sectors;
> +
> +    return 0;
> +}
> +
>  static int qcow2_set_key(BlockDriverState *bs, const char *key)
>  {
>      BDRVQcowState *s = bs->opaque;
> @@ -2268,6 +2276,7 @@ static BlockDriver bdrv_qcow2 = {
>  
>      .bdrv_change_backing_file   = qcow2_change_backing_file,
>  
> +    .bdrv_refresh_limits        = qcow2_refresh_limits,
>      .bdrv_invalidate_cache      = qcow2_invalidate_cache,
>  
>      .create_options = qcow2_create_options,
> diff --git a/block/qed.c b/block/qed.c
> index 450a1fa..d35bd1c 100644
> --- a/block/qed.c
> +++ b/block/qed.c
> @@ -495,7 +495,6 @@ static int bdrv_qed_open(BlockDriverState *bs, QDict 
> *options, int flags,
>          }
>      }
>  
> -    bs->bl.write_zeroes_alignment = s->header.cluster_size >> 
> BDRV_SECTOR_BITS;
>      s->need_check_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
>                                              qed_need_check_timer_cb, s);
>  
> @@ -507,6 +506,15 @@ out:
>      return ret;
>  }
>  
> +static int bdrv_qed_refresh_limits(BlockDriverState *bs)
> +{
> +    BDRVQEDState *s = bs->opaque;
> +
> +    bs->bl.write_zeroes_alignment = s->header.cluster_size >> 
> BDRV_SECTOR_BITS;
> +
> +    return 0;
> +}
> +
>  /* We have nothing to do for QED reopen, stubs just return
>   * success */
>  static int bdrv_qed_reopen_prepare(BDRVReopenState *state,
> @@ -1616,6 +1624,7 @@ static BlockDriver bdrv_qed = {
>      .bdrv_truncate            = bdrv_qed_truncate,
>      .bdrv_getlength           = bdrv_qed_getlength,
>      .bdrv_get_info            = bdrv_qed_get_info,
> +    .bdrv_refresh_limits      = bdrv_qed_refresh_limits,
>      .bdrv_change_backing_file = bdrv_qed_change_backing_file,
>      .bdrv_invalidate_cache    = bdrv_qed_invalidate_cache,
>      .bdrv_check               = bdrv_qed_check,
> diff --git a/block/vmdk.c b/block/vmdk.c
> index 0734bc2..f8a387f 100644
> --- a/block/vmdk.c
> +++ b/block/vmdk.c
> @@ -428,10 +428,6 @@ static int vmdk_add_extent(BlockDriverState *bs,
>      extent->l2_size = l2_size;
>      extent->cluster_sectors = flat ? sectors : cluster_sectors;
>  
> -    if (!flat) {
> -        bs->bl.write_zeroes_alignment =
> -            MAX(bs->bl.write_zeroes_alignment, cluster_sectors);
> -    }
>      if (s->num_extents > 1) {
>          extent->end_sector = (*(extent - 1)).end_sector + extent->sectors;
>      } else {
> @@ -886,6 +882,23 @@ fail:
>      return ret;
>  }
>  
> +
> +static int vmdk_refresh_limits(BlockDriverState *bs)
> +{
> +    BDRVVmdkState *s = bs->opaque;
> +    int i;
> +
> +    for (i = 0; i < s->num_extents; i++) {
> +        if (!s->extents[i].flat) {
> +            bs->bl.write_zeroes_alignment =
> +                MAX(bs->bl.write_zeroes_alignment,
> +                    s->extents[i].cluster_sectors);
> +        }
> +    }
> +
> +    return 0;
> +}
> +
>  static int get_whole_cluster(BlockDriverState *bs,
>                  VmdkExtent *extent,
>                  uint64_t cluster_offset,
> @@ -1971,6 +1984,7 @@ static BlockDriver bdrv_vmdk = {
>      .bdrv_get_allocated_file_size = vmdk_get_allocated_file_size,
>      .bdrv_has_zero_init           = vmdk_has_zero_init,
>      .bdrv_get_specific_info       = vmdk_get_specific_info,
> +    .bdrv_refresh_limits          = vmdk_refresh_limits,
>  
>      .create_options               = vmdk_create_options,
>  };
> diff --git a/include/block/block_int.h b/include/block/block_int.h
> index 8b132d7..c49fa6b 100644
> --- a/include/block/block_int.h
> +++ b/include/block/block_int.h
> @@ -226,6 +226,8 @@ struct BlockDriver {
>      int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
>      bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
>  
> +    int (*bdrv_refresh_limits)(BlockDriverState *bs);
> +
>      /*
>       * Returns 1 if newly created images are guaranteed to contain only
>       * zeros, 0 otherwise.