Sorry for resending (fixed a typo on prev mail) - Can someone please reply to this? On Tue, Jun 29, 2010 at 6:27 PM, shailesh parulekar <shailesh.parulekar@xxxxxxxxx> wrote: > Hi, > > I am writing a SCSI LLD and have a query regarding handling of underflow errors. > > In function sd_prep_fn (sd.c) the code populates scsi command pointer > (SCpnt) and it's cmnd. > Going ahead it also sets the total length to be transferred in the > scsi data buffer (SCpnt->sdb.length = this_count * sdp->sector_size). > Here 'this_count' looks to be the nr_sectors. This assignment looks > fine too. > But later when assigning the 'underflow' field of SCpnt, the > assignment uses a hard-coded bit shift (SCpnt->underflow = this_count > << 9). Is this valid? OR > should it be: SCpnt->underflow = this_count * sdp->sector_size? > > So, if the sector size exported by the target device is greater than > 512 (4K in our case), the SCpnt->underflow is not same as > SCpnt->sdb.length. The later is 8 times bigger. So if I get a read of > say 8K the 'underflow' is set only to 1K. Now, if the target say > returns only 2K of data for this read, the amount of data transferred > is not less than residual (1K in this case). So is the LLD not > supposed to return (DID_ERROR << 16) in 'result' for this case > (assuming target returns result == GOOD)? > > Please let me know. > > thanks, > --shailesh > -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
