| Message ID | 1466751651-9900-3-git-send-email-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
>>>>> "Christoph" == Christoph Hellwig <hch@lst.de> writes: Christoph> WRITE SAME is a data integrity operation and we can't simply Christoph> ignore errors. Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
On Wed, Jun 29, 2016 at 01:04:02AM -0400, Martin K. Petersen wrote: > >>>>> "Christoph" == Christoph Hellwig <hch@lst.de> writes: > > Christoph> WRITE SAME is a data integrity operation and we can't simply > Christoph> ignore errors. > > Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> > > -- > Martin K. Petersen Oracle Linux Engineering Along with the first patch this resolves the problems I was seeing when using the script below when applied atop of axboe/linux-block.git branch for-4.8/core (commit e118e4be8a96c6b89307ee3d360d954249981b8f). The only snag is that when userland manually disables max_write_same_blocks on the base device (by echoing 0 into it) stacked devices will not set their their own queue/write_same_max_bytes to 0 when the BLKZEROOUT fails to use write same (but they do correctly zero the device). If the base device somehow internally toggles this to 0 itself during a write same then dm devices seem to set queue/write_same_max_bytes to 0. Tested-by: Sitsofe Wheeler <sitsofe@yahoo.com> #!/bin/bash # Check BLKZEROOUT behaviour modprobe scsi_debug write_same_length=131072 dev=$(lsblk -n -p -S | awk '/scsi_debug/ { print $1 }') dev_kname=$(lsblk -n -S ${dev} | awk '{ print $1 }') sys_dev_mwsb_path=$(realpath /sys/block/"${dev_kname}"/device/scsi_disk/*/max_write_same_blocks) pvcreate ${dev} vgcreate vg ${dev} lvcreate -L4M -n lv vg lv_dev=/dev/mapper/vg-lv sys_lv_wsmb_path="/sys/block/$(lsblk -n -o KNAME "${lv_dev}")/queue/write_same_max_bytes" orig_max_ws=$(cat "${sys_dev_mwsb_path}") echo "Stacked device's write_same_max_bytes prior blkzeroout: $(cat ${sys_lv_wsmb_path})" echo "Forcing base device's max_write_same_blocks to 0" echo 0 > "${sys_dev_mwsb_path}" size=$(blockdev --getsize64 "${lv_dev}") tr '\0' '\377' < /dev/zero | dd oflag=direct bs=64K of="${lv_dev}" count=$((${size} / (64 * 1024))) status=none blkdiscard -v --zero "${lv_dev}" dd if="${lv_dev}" iflag=direct bs=64K status=none | cmp -b --bytes "${size}" /dev/zero - verification=$? echo "Stacked device's write_same_max_bytes after blkzeroout: $(cat ${sys_lv_wsmb_path})" lvchange -a n "${lv_dev}"
diff --git a/block/blk-lib.c b/block/blk-lib.c index 45b35b1..e371f83 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -178,7 +178,7 @@ int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, if (bio) ret = submit_bio_wait(bio); - return ret != -EOPNOTSUPP ? ret : 0; + return ret; } EXPORT_SYMBOL(blkdev_issue_write_same);
WRITE SAME is a data integrity operation and we can't simply ignore errors. Signed-off-by: Christoph Hellwig <hch@lst.de> --- block/blk-lib.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)