Re: [Qemu-devel] [PATCH] block: init bs->io_base correctly to avoid locking

[Paolo Bonzini <pbonzini@redhat.com>]

Il 10/01/2013 10:21, Peter Lieven ha scritto:
> If io_limits are specified during runtime that exceed the number of
> operations in flight
> bs->io_base is not initialized in the else statement in
> bdrv_exceed_io_limits().
> The wait time calculated in bdrv_exceed_{bps,iops}_limits is thus
> totally wrong
> and the machine locks.
> 
> Signed-off-by: Peter Lieven <pl@kamp.de>
> ---
>  block.c |    4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/block.c b/block.c
> index 4e28c55..309aa85 100644
> --- a/block.c
> +++ b/block.c
> @@ -159,6 +159,10 @@ void bdrv_io_limits_enable(BlockDriverState *bs)
>      bs->slice_start = qemu_get_clock_ns(vm_clock);
>      bs->slice_end   = bs->slice_start + bs->slice_time;
>      memset(&bs->io_base, 0, sizeof(bs->io_base));

Please remove this memset.

> +    bs->io_base.bytes[0] = bs->nr_bytes[0];
> +    bs->io_base.bytes[1] = bs->nr_bytes[1];
> +    bs->io_base.ios[0] = bs->nr_ops[0];
> +    bs->io_base.ios[1] = bs->nr_ops[1];
>      bs->io_limits_enabled = true;
>  }
> 

Also, perhaps you can just call

    bdrv_exceed_io_limits(bs, 0, 0, NULL);

(which also subsumes the setting of slice_time, slice_start, slice_end).

Paolo