Re: drivers/block/ub.c

[Alan Stern]

On Sun, 27 Jun 2004, Oliver Neukum wrote:

> Am Samstag, 26. Juni 2004 22:06 schrieb Pete Zaitcev:
> > +static int ub_submit_top_sense(struct ub_dev *sc, struct ub_scsi_cmd *cmd)
> > +{
> > +       struct ub_scsi_cmd *scmd;
> > +
> > +       scmd = &sc->top_rqs_cmd;
> > +
> > +       /* XXX Can be done at init */
> > +       scmd->cdb[0] = REQUEST_SENSE;
> > +       scmd->cdb_len = 6;
> > +       scmd->dir = UB_DIR_READ;
> > +       scmd->state = UB_CMDST_INIT;
> > +       scmd->data = sc->top_sense;
> 
> You must allocate a separate buffer to the sense data.
> We had similar code in hid which leads to data corruption
> on some architectures. It's an issue of DMA coherency.

I mentioned this some time ago to the SCSI maintainers.  They felt that it 
wasn't necessary to allocate a separate buffer for the sense data -- I'm 
not sure why.  Apparently most if not all SCSI drivers fail to do this.  
Usb-storage doesn't do it either; maybe we should.

Alan Stern

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/