> -----Original Message----- > From: Martin K. Petersen [mailto:martin.petersen@xxxxxxxxxx] > Sent: Thursday, February 22, 2018 9:52 AM > To: Shivasharan S <shivasharan.srikanteshwara@xxxxxxxxxxxx> > Cc: linux-scsi@xxxxxxxxxxxxxxx; kashyap.desai@xxxxxxxxxxxx; > sumit.saxena@xxxxxxxxxxxx > Subject: Re: [PATCH] megaraid_sas: Re-enable WRITE SAME > > > Hi Shivasharan! > > > For MegaRAID controller we want to support WRITE SAME for non-raid > > volumes. Because of host wide "no_write_same = 1" settings, we are > > not able to use provision mode. > > The no_write_same flag gates whether the WRITE SAME(10/16) commands > are > used to zero block ranges. It does not influence whether WRITE SAME with > the UNMAP bit set is used to deallocate block ranges on devices which > support logical block provisioning. Hi Martin, Thanks for this clarification. I checked further to understand difference between zeroout block ranges and deallocation (unmap) of block ranges using WRITE SAME command and impact of setting no_write_same flag. So my understanding is when driver sets no_write_same flag, it is only disabling BLKZEROOUT/write_zeroes requests from using WRITE SAME command. Instead individual writes are sent to zero the blocks (if BLKDEV_ZERO_NOFALLBACK flag is not set). > > > Virtual Disk will not enable WRITE SAME feature providing correct VPD > > block limit page. > > This wasn't always the case. > > If you want to enable WRITE SAME you'll have to provide some sort of > heuristic to avoid breaking things on firmware versions and controllers > that predate correct reporting. > Our inhouse driver code did not set the no_write_same flag. And so far we have not seen any issues in this area with inhouse driver. I am running tests now to check firmware behavior you mentioned. Will get back to you with the results. Are you aware of any issue that led to this change? > > Virtual Disk in MegaRAID provides below values in VPD page 0xb0 which > > will disable WRITE SAME commands. > > A value of 0 in > > Maximum write same length: 0x0 blocks > > means the maximum length is not reported. It does not mean that WRITE > SAME isn't supported. > Ok. With latest firmware, for virtual disks we do not set write same/unmap support in the logical block provisioning VPD page. Will check behavior with older firmware versions also and get back. Logical block provisioning VPD page (SBC): Unmap command supported (LBPU): 0 Write same (16) with unmap bit supported (LBWS): 0 Write same (10) with unmap bit supported (LBWS10): 0 Logical block provisioning read zeros (LBPRZ): 0 Anchored LBAs supported (ANC_SUP): 0 Threshold exponent: 0 Descriptor present (DP): 0 Minimum percentage: 0 Provisioning type: 0 Threshold percentage: 0 Thanks, Shivasharan > -- > Martin K. Petersen Oracle Linux Engineering
