Damien,
> + if (sdkp->max_ws_blocks &&
> + sdkp->physical_block_size > logical_block_size) {
> + /*
> + * Reporting a maximum number of blocks that is not aligned
> + * on the device physical size would cause a large write same
> + * request to be split into physically unaligned chunks by
> + * __blkdev_issue_write_zeroes() and __blkdev_issue_write_same()
> + * even if the caller of these functions took care to align the
> + * large request. So make sure the maximum reported is aligned
> + * to the device physical block size. This is only an optional
> + * optimization for regular disks, but this is mandatory to
> + * avoid failure of large write same requests directed at
> + * sequential write required zones of host-managed ZBC disks.
> + */
> + sector_t phys_mask =
> + bytes_to_logical(sdkp->device,
> + sdkp->physical_block_size) - 1;
> +
> + sdkp->max_ws_blocks &= ~phys_mask;
> + }
> +
> out:
> blk_queue_max_write_same_sectors(q, sdkp->max_ws_blocks *
> (logical_block_size >> 9));
ALIGN_DOWN(sdkp->max_ws_blocks, sdkp->physical_block_size)?
--
Martin K. Petersen Oracle Linux Engineering
