On 5/31/21 9:04 AM, Bean Huo wrote: > I did a UFS queue limitation test before, observed that once the > queue is full, then the active task number in the queue will get > down. For the Nvme, the scenario is the same. You can refer to the > slide 23, and slide 24 in the pdf: > https://elinux.org/images/6/6c/Linux_Storage_System_Bottleneck_Exploration_V0.3.pdf > I don't know if your patch can fix this issue. Hi Bean, That's a very interesting presentation. Unfortunately the overhead for SCSI in that presentation includes the SCSI core + UFS driver. Given the use of the host lock in the version of the UFS driver that was used to prepare that presentation, the overhead introduced by the UFS driver may be significant. Maybe someone should measure the overhead of the scsi_debug driver and compare it with the NVMe loopback driver to get a better idea of how the overhead of these two subsystems compare to each other? Bart.
