block: Add bdrv_has_zero_init_truncate()
commitceaca56feee6a1d682c882300855cad44a67ec8e
authorMax Reitz <mreitz@redhat.com>
Wed, 24 Jul 2019 17:12:31 +0000 (24 19:12 +0200)
committerMax Reitz <mreitz@redhat.com>
Mon, 19 Aug 2019 15:13:26 +0000 (19 17:13 +0200)
tree28fe6b35daa9996adbc486fe57b7c606ff602fdc
parentcdf3bc934ad1e5319b03f2c85f381f5ffd2f8ca8
block: Add bdrv_has_zero_init_truncate()

No .bdrv_has_zero_init() implementation returns 1 if growing the file
would add non-zero areas (at least with PREALLOC_MODE_OFF), so using it
in lieu of this new function was always safe.

But on the other hand, it is possible that growing an image that is not
zero-initialized would still add a zero-initialized area, like when
using nonpreallocating truncation on a preallocated image.  For callers
that care only about truncation, not about creation with potential
preallocation, this new function is useful.

Alternatively, we could have added a PreallocMode parameter to
bdrv_has_zero_init().  But the only user would have been qemu-img
convert, which does not have a plain PreallocMode value right now -- it
would have to parse the creation option to obtain it.  Therefore, the
simpler solution is to let bdrv_has_zero_init() inquire the
preallocation status and add the new bdrv_has_zero_init_truncate() that
presupposes PREALLOC_MODE_OFF.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20190724171239.8764-4-mreitz@redhat.com
Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
block.c
include/block/block.h
include/block/block_int.h