On 5/16/24 18:44, Wu Bo wrote:
So I figured this solution to fix this priority issue on zoned device. It sure raises the overhead but can do fix it.
Something I should have realized earlier is that this patch is not necessary with the latest upstream kernel (v6.10-rc1). Damien's zoned write plugging patch series has been merged. Hence, I/O schedulers, including the mq-deadline I/O schedulers, will only see a single zoned write at a time per zone. So it is no longer possible that zoned writes are reordered by the I/O scheduler because of their I/O priorities. Thanks, Bart.
