In general, writes to SCSI disks are required to align to the logical block size. On the other hand, ZBC and ZAC specifications require that the writes to sequential write required zones to align to the physical block size. When ZBC/ZAC host-managed zoned disks have the physical block size different from the logical block size, writes aligned to the logical block size fail. The sysfs attribute zone_write_granularity was introduced so that userland programs can tell what is the alignment size for sequential write required zones. As for ZBC/ZAC host-managed zoned disks, zone_write_granularity shows the physical block size. However, there are two issues related to this requirement of the physical block size alignment. The first issue is unclear failure report. On the write fail, the disks return the unaligned write command error, which may happen for other causes other than the physical block size alignment. The second issue is wrong value of zone_write_granularity sysfs attribute. In most cases, it shows correct values. But during revalidate of the disks, it shows wrong values. The two patches in this series address the two issues. Shin'ichiro Kawasaki (2): scsi: sd: Check physical sector alignment of sequential zone writes scsi: sd: Fix wrong zone_write_granularity value at revalidate drivers/scsi/sd.c | 17 ++++++++++++++++- drivers/scsi/sd_zbc.c | 8 -------- 2 files changed, 16 insertions(+), 9 deletions(-) -- 2.38.1
