Bart, Thanks a lot, Avri +AD4- -----Original Message----- +AD4- From: Bart Van Assche +AD4- Sent: Wednesday, August 01, 2018 6:13 PM +AD4- To: hch+AEA-lst.de+ADs- Avri Altman +ADs- linux-scsi+AEA-vger.kernel.org+ADs- +AD4- jthumshirn+AEA-suse.de+ADs- hare+AEA-suse.com+ADs- martin.petersen+AEA-oracle.com+ADs- +AD4- jejb+AEA-linux.vnet.ibm.com +AD4- Cc: Vinayak Holikatti +ADs- Avi Shchislowski +ADs- Alex Lemberg +ADs- Stanislav Nijnikov +ADs- +AD4- subhashj+AEA-codeaurora.org +AD4- Subject: Re: +AFs-PATCH 2/6+AF0- scsi: ufs: Add ufs-bsg module +AD4- +AD4- On Wed, 2018-08-01 at 11:04 +-0300, Avri Altman wrote: +AD4- +AD4- +-+-+- b/drivers/scsi/ufs/ufs+AF8-bsg.c +AD4- +AD4- +AEAAQA- -0,0 +-1,127 +AEAAQA- +AD4- +AD4- +-// SPDX-License-Identifier: GPL-2.0 +AD4- +AD4- +-/+ACo- +AD4- +AD4- +- +ACo- SSCSI transport companion for UFS HBA +AD4- +AD4- What is +ACI-SSCSI+ACI-? Done +AD4- +AD4- +AD4- diff --git a/drivers/scsi/ufs/ufs+AF8-bsg.h b/drivers/scsi/ufs/ufs+AF8-bsg.h +AD4- +AD4- new file mode 100644 +AD4- +AD4- +AFs- ... +AF0- +AD4- +AD4- +AD4- +-struct ufs+AF8-bsg+AF8-request +AHs- +AD4- +AD4- +- uint32+AF8-t msgcode+ADs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-header header+ADs- +AD4- +AD4- +- union +AHs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-query qr+ADs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-task+AF8-req tr+ADs- +AD4- +AD4- +- /+ACo- use utp+AF8-upiu+AF8-query to host the 4 dwards of uic command +AD4- +ACo-/ +AD4- +AD4- What are +ACI-dwards+ACI-? Done. +AD4- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-query uc+ADs- +AD4- +AD4- +- +AH0- tsf+ADs- +AD4- +AD4- +- uint8+AF8-t data+AFs-0+AF0AOw- +AD4- +AD4- The offset of the data member will change if a member will be added to the +AD4- union with a larger size than the existing members. That seems like an API +AD4- design bug to me. Actually I am not using it, so will just remove it. +AD4- +AD4- +AD4- +-+AH0- +AF8AXw-packed+ADs- +AD4- +AD4- Would the data member offsets be the same without +ACIAXwBf-packed+ACI-? If so, +AD4- +AF8AXw-packed +AD4- should be left out because it results in generation of suboptimal code on +AD4- architectures that do not support unaligned multi-byte reads (e.g. IA-64). Done +AD4- +AD4- +AD4- +-struct ufs+AF8-bsg+AF8-reply +AHs- +AD4- +AD4- +- /+ACo- +AD4- +AD4- +- +ACo- The completion result. Result exists in two forms: +AD4- +AD4- +- +ACo- if negative, it is an -Exxx system errno value. There will +AD4- +AD4- +- +ACo- be no further reply information supplied. +AD4- +AD4- +- +ACo- else, it's the 4-byte scsi error result, with driver, host, +AD4- +AD4- +- +ACo- msg and status fields. The per-msgcode reply structure +AD4- +AD4- +- +ACo- will contain valid data. +AD4- +AD4- +- +ACo-/ +AD4- +AD4- +- uint32+AF8-t result+ADs- +AD4- +AD4- +- +AD4- +AD4- +- /+ACo- If there was reply+AF8-payload, how much was recevied ? +ACo-/ +AD4- +AF4AXgBeAF4AXgBeAF4AXg- +AD4- Did you perhaps mean +ACI-received+ACI-? Done. +AD4- +AD4- +AD4- +- uint32+AF8-t reply+AF8-payload+AF8-rcv+AF8-len+ADs- +AD4- +AD4- +- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-header header+ADs- +AD4- +AD4- +- union +AHs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-query qr+ADs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-task+AF8-rsp tr+ADs- +AD4- +AD4- +- struct utp+AF8-upiu+AF8-query uc+ADs- +AD4- +AD4- +- +AH0- tsf+ADs- +AD4- +AD4- +- uint8+AF8-t data+AFs-0+AF0AOw- +AD4- +AD4- +-+AH0AOw- +AD4- +AD4- +- +AD4- +AD4- +-struct ufs+AF8-bsg+AF8-raw+AF8-upiu +AHs- +AD4- +AD4- +- struct ufs+AF8-bsg+AF8-upiu request+ADs- +AD4- +AD4- +- struct ufs+AF8-bsg+AF8-upiu reply+ADs- +AD4- +AD4- +-+AH0AOw- +AD4- +AD4- Are any of the above data structures needed by user space software? Should +AD4- these +AD4- data structure definitions perhaps be moved to a header file under +AD4- include/uapi? I wanted to benefit from ufs.h and ufshcd.h so used it here. In include/uapi/scsi/scsi+AF8-bsg+AF8-ufs.h I defined an 8 dw+ACo-o+ACo-rds opaque structs for the request and reply. +AD4- +AD4- Thanks, +AD4- +AD4- Bart.
