On 2019/11/05 14:19, Martin K. Petersen wrote: > > Damien, > >> The SG driver can make use of this field to keep the io header resid >> as an int, with negative values indicating overflows and positive >> values underflows. > > I am all for synthesizing what SG returns to userland. > > That is also the case in the context of Hannes' SCSI result revamp. I > would much prefer to have well-defined and consistent internal kernel > status fields and then transmogrify those into something compatible with > what userland applications might expect. As opposed to perpetuating the > train wreck that is the current scsi_cmnd result. OK. But just to clarify, do you mean changing struct scsi_request resid_len field to an int and considering positive values of it as underflow and negative values as overflow ? Or keeping resid_len as an unsigned int and adding a flag specifying if the value means underflow or overflow ? I am rely surprised that my simple patch resulted in such a big discussion. But if that leads to code improvements, then let's drop it and work on a new series ! -- Damien Le Moal Western Digital Research
