[sheepdog] [PATCH v3 03/17] block/io: use int64_t bytes parameter in bdrv_check_byte_request()

Alberto Garcia berto at igalia.com
Mon May 11 17:57:15 CEST 2020


On Thu 30 Apr 2020 01:10:19 PM CEST, Vladimir Sementsov-Ogievskiy wrote:
> We are generally moving to int64_t for both offset and bytes parameters
> on all io paths.
>
> Main motivation is realization of 64-bit write_zeroes operation for
> fast zeroing large disk chunks, up to the whole disk.
>
> We chose signed type, to be consistent with off_t (which is signed) and
> with possibility for signed return type (where negative value means
> error).
>
> So, convert bdrv_check_byte_request() now.
>
> Patch-correctness audit by Eric Blake:
>
>   This changes an unsigned to signed value on 64-bit machines, and
>   additionally widens the parameter on 32-bit machines.  Existing
>   callers:
>
>   bdrv_co_preadv_part() with 'unsigned int' - no impact
>   bdrv_co_pwritev_part() with 'unsigned int' - no impact
>   bdrv_co_copy_range_internal() with 'uint64_t' -
>       potentially fixes a latent bug on 32-bit machines. Requires a
>       larger audit to see how bdrv_co_copy_range() and friends are
>       used:
>
>   block/block-backend.c:blk_co_copy_range() - passes 'int', thus < 2G
>   block/block-copy.c:block_copy_do_copy() -
>       passes 'int64_t', but only after assert(nbytes < INT_MAX); also
>       it has a BLOCK_COPY_MAX_COPY_RANGE set to 16M that factors into
>       its calculations on how much to do per iteration
>
>   So it looks like at present we are safe.
>
> Series: 64bit-block-status
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov at virtuozzo.com>
> Reviewed-by: Eric Blake <eblake at redhat.com>

Reviewed-by: Alberto Garcia <berto at igalia.com>

Berto


More information about the sheepdog mailing list