Kirill, > One more thing to discuss. The new REQ_NOZERO flag won't be supported > by many block devices (their number will be even less, than number of > REQ_OP_WRITE_ZEROES supporters). Will this be a good thing, in case of > we will be completing BLKDEV_ZERO_ALLOCATE bios in > __blkdev_issue_write_zeroes() before splitting? I mean introduction of > some flag in struct request_queue::limits. Completion of them with > -EOPNOTSUPP in block devices drivers looks suboptimal for me. We already have the NOFALLBACK flag to let the user make that decision. If that flag is not specified, and I receive an allocate request for a SCSI device that does not support ANCHOR, my expectation would be that I would do a regular write same. If it's a filesystem that is the recipient of the operation and not a SCSI device, how to react would depend on how the filesystem handles unwritten extents, etc. -- Martin K. Petersen Oracle Linux Engineering
