On Wed, 2017-05-10 at 20:31 +0000, Bart Van Assche wrote: > On Tue, 2017-05-09 at 21:03 -0700, Nicholas A. Bellinger wrote: > > In any event, the point is your patch to add sbc_parse_verify() broke > > existing behavior of WRITE_VERIFY_* by dropping SCF_SCSI_DATA_CDB > > assignment for all cases. > > As I had already explained in detail I disagree with this statement. BTW, did > you know that your patch "target: Fix sbc_parse_verify bytchk = 0 handling" is > not sufficient to avoid a buffer overflow in the iSCSI target driver? One way > to trigger a buffer overflow is by making the initiator send more immediate > data than the Data-Out buffer size derived from the CDB. > If you think you've found a legitimate bug, then post the test case to trigger it atop what's in target-pending/for-next for -rc1. Regardless, your WRITE_VERIFY changes broke existing behavior, and I'm not going to merge a bunch of new code at the last minute.