Re: [PATCH 2/3] block/file-posix: consider file size when fallback to max_align

[Max Reitz]

On 30.01.20 16:22, Vladimir Sementsov-Ogievskiy wrote:
> If we failed to probe request_align, we fallback to max_align. But
> this is wrong, if file size is not aligned to our max_align. Let's
> instead chose alignment so that file size is a multiple of it.

It’s entirely possible that the file size is not aligned to the request
alignment, though.  If so, this patch will make the whole file inaccessible.

> Signed-off-by: Vladimir Sementsov-Ogievskiy <address@hidden>
> ---
>  block/file-posix.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/block/file-posix.c b/block/file-posix.c
> index 7f366046c2..e9c4e509f6 100644
> --- a/block/file-posix.c
> +++ b/block/file-posix.c
> @@ -385,11 +385,25 @@ static void raw_probe_alignment(BlockDriverState *bs, 
> int fd, Error **errp)
>              align = alignments[i];
>              if (raw_is_io_aligned(fd, buf, align)) {
>                  /* Fallback to safe value. */
> -                bs->bl.request_alignment = (align != 1) ? align : max_align;
> +                bs->bl.request_alignment = align;

The comment above is now outdated.

Max

>                  break;
>              }
>          }
>          qemu_vfree(buf);
> +
> +        if (bs->bl.request_alignment == 1) {
> +            /*
> +             * Succeed to read with alignment = 1. But it may be unallocated
> +             * area on XFS, and we'll fail later if keep request_alignment = 
> 1.
> +             *
> +             * Chose safer alignment, keeping in mind file size if possible.
> +             */
> +
> +            int64_t len = raw_getlength_fd(bs, fd);
> +
> +            bs->bl.request_alignment =
> +                    len <= 0 ? max_align : MIN(max_align, len & ~(len - 1));
> +        }
>      }
>  
>      if (!s->buf_align) {
>

signature.asc
Description: OpenPGP digital signature