On Tue, Jun 08, 2021 at 03:16:30PM +0200, Paolo Bonzini wrote:
> For block host devices, I/O can happen through either the kernel file
> descriptor I/O system calls (preadv/pwritev, io_submit, io_uring)
> or the SCSI passthrough ioctl SG_IO.
>
> In the latter case, the size of each transfer can be limited by the
> HBA, while for file descriptor I/O the kernel is able to split and
> merge I/O in smaller pieces as needed. Applying the HBA limits to
> file descriptor I/O results in more system calls and suboptimal
> performance, so this patch splits the max_transfer limit in two:
> max_transfer remains valid and is used in general, while max_hw_transfer
> is limited to the maximum hardware size. max_hw_transfer can then be
> included by the scsi-generic driver in the block limits page, to ensure
> that the stricter hardware limit is used.
>
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> block/block-backend.c | 12 ++++++++++++
> block/file-posix.c | 2 +-
> block/io.c | 1 +
> hw/scsi/scsi-generic.c | 2 +-
> include/block/block_int.h | 7 +++++++
> include/sysemu/block-backend.h | 1 +
[you can use git's orderfile option to put .h changes first]
> 6 files changed, 23 insertions(+), 2 deletions(-)
>
> diff --git a/block/block-backend.c b/block/block-backend.c
> index 15f1ea4288..2ea1412a54 100644
> --- a/block/block-backend.c
> +++ b/block/block-backend.c
> @@ -1953,6 +1953,18 @@ uint32_t blk_get_request_alignment(BlockBackend *blk)
> return bs ? bs->bl.request_alignment : BDRV_SECTOR_SIZE;
> }
>
> +/* Returns the maximum hardware transfer length, in bytes; guaranteed
> nonzero */
> +uint64_t blk_get_max_hw_transfer(BlockBackend *blk)
Since we have declared (elsewhere) that the maximum block device is
signed, would this be better as int64_t? (Our reasoning is that off_t
is also signed, and we are unlikely to need to handle anything bigger
than what off_t can handle; plus it leaves room for returning errors,
although this particular function is not giving errors; see also
Vladimir's work on making the block layer 64-bit clean). I'm not
opposed to unsigned here to represent lack of errors, but consistency
with the rest of the block layer may argue for signed.
> +++ b/include/block/block_int.h
> @@ -695,6 +695,13 @@ typedef struct BlockLimits {
> * clamped down. */
> uint32_t max_transfer;
>
> + /* Maximal hardware transfer length in bytes. Applies whenever
> + * transfers to the device bypass the kernel I/O scheduler, for
> + * example with SG_IO. If larger than max_transfer or if zero,
> + * blk_get_max_hw_transfer will fall back to max_transfer.
> + */
> + uint64_t max_hw_transfer;
> +
Reviewed-by: Eric Blake <[email protected]>
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
