On Wed, 2021-07-28 at 10:39 +0900, Chanho Park wrote: > > Hi Chanho Park, > > Thansk for yoru reply. > > I didn't see your changes about task_tag and IID. Having a look at > > ufshcd_prepare_utp_scsi_cmd_upiu(), the task tag in the UPIU header > > is > > still only task tag. and IID is always 0x00. > > If you didn't add these changes, your patch is un-readable, and > > also the > > driver doesn't have a real usage case. > > > I already replied regarding this in another mail thread and please > refer below comments. The controller will handle the tag translation. > > https://lore.kernel.org/linux-scsi/002901d782c6$937ac0f0$ba7042d0$@samsung.com/ > > > > > Also, you mentioned there is no support/change needed from the UFS > > device > > side. But, IMO, if you changed the UPIU header, there are changes > > needed > > on the UFS device side in order to use your driver. > > > Please see the figure[1]. Once IID_IN_TASKTAG bit is set, the > function arbiter will translate the value of UPIU header before > delivering it to the device. > > > > [1]: > https://lore.kernel.org/linux-scsi/20210714071131.101204-1-chanho61.park@xxxxxxxxxxx/ > > > > Best Regards, > > Chanho Park Hi Chanho Park, thansk for your further reply. I didn't find out your cover-letter in your patchset before. IID will be automatically set by HW controller according to the VH_ID in the tasg_tag byte[7:5]. Before sending this RPMB to UFS device, doese your controller reset task_tag byte[7:5] to 0x0 or your controller will keep VH_ID? one more question, you mentioned para-virtualization, did you use Xen or KVM? Kind regards, Bean
