I am really starting to empathize with Andre here. He's talking safety,
you guys preach security; he says via the kernel API, you say there is
still RAW_IO. The problem is that these commands are VENDOR SPECIFIC,
outside the ATA-ATAPI specs and undocumented/not publicly available. As
such you don't KNOW if you're sending disktobrick to the drive or not.
What Andre wants to do is limit the kernel API to the ATA-ATAPI specs so
this cannot happen in normal/development use. If you're worried about
losing the ability to FLASH your firmware, just use an SUID root userland
program which writes via RAW-IO. At least in this way the program will
know what these commands are and use them in the proper context; not frying
your hardware when testing a new driver for robustness ie. properly
returning -ENOTATA instead.
just my $.02 cents worth - seems everybody else got there's. :)
--- Tim
Oliver Xymoron wrote:
>
> On Sat, 22 Jul 2000, Vojtech Pavlik wrote:
>
> > On Fri, Jul 21, 2000 at 08:25:06PM -0500, Oliver Xymoron wrote:
> > > On Fri, 21 Jul 2000, Andre Hedrick wrote:
> > >
> > > > With out the full touch it will not work.
> > > > 20% are functionally that have to be there for 2.4 to stand a chance of
> > > > being correct.
> > >
> > > Ok, separate out those parts and submit them on their own. I read your
> > > patch and I figured that some percentage was clearly unrelated.
> > >
> > > > 80% is TASKFILE rewrite to give us the armor.
> > >
> > > That means still 50k and that's way too large a rewrite to be
> > > considered a bugfix. I think something like the below can do it in less
> > > than 10 lines for 2.4 and might even pass the "obviously correct"
> > > criterion. Save the command parser business for 2.5.
> >
> > Ok, what about just this simple little patch?
>
> Aside from having lines longer than 80 columns it looks good. Here's the
> analogous SCSI patch:
>
> --- scsi_ioctl.c~ Tue Mar 14 00:15:03 2000
> +++ scsi_ioctl.c Sat Jul 22 11:21:36 2000
> @@ -410,7 +410,7 @@
> case SCSI_IOCTL_PROBE_HOST:
> return ioctl_probe(dev->host, arg);
> case SCSI_IOCTL_SEND_COMMAND:
> - if (!capable(CAP_SYS_ADMIN))
> + if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
> return -EACCES;
> return scsi_ioctl_send_command((Scsi_Device *) dev,
> (Scsi_Ioctl_Command *) arg);
>
> --
> "Love the dolphins," she advised him. "Write by W.A.S.T.E.."
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Jul 23 2000 - 21:00:19 EST