--------- Received message begins Here ---------
Tim Waugh <[EMAIL PROTECTED]>:
> On Thu, Jan 04, 2001 at 03:39:10PM +0100, Andrea Arcangeli wrote:
>
> > As noted yesterday falling into parport_write will silenty lose data when the
> > printer is off.
>
> (Actually it depends; I think FIFO/DMA paths are fine, but yes, the
> software implementation can lose data.)
>
> > If it's not feasible to make parport_write reliable against
> > power-off printer, then I recommend to loop in interruptible mode
> > before entering the main loop (waiting the printer to power-on) like
> > in latest patch from Peter.
>
> Have I missed a patch? How do you know whether or not the printer is
> on yet?
>
> As I understand it, you can't guarantee anything about any of the
> signals when the printer is off, so all you can do is look for
> 'suspicous' things (like 'no error' and 'paper out'). But some
> printers do this during normal operation, and hence the LP_CAREFUL
> switch.
>
> Return -EIO when the printer is on and off-line is a bug, sure enough.
> That's what the -EAGAIN patch was for, and Peter's patch fixes this
> too.
>
> But if you want to avoid losing data when your printer is off you need
> to use LP_CAREFUL, and hope printing still works at all (depends on
> your printer).
>
> If this goes away:
>
> if ((status & LP_PERRORP) && !(LP_F(minor) & LP_CAREFUL))
> /* No error. */
> last = 0;
>
> then some people might not be able to print at all.
I don't believe this is a problem that CAN be fixe, either by hardware
or software.
Originally, (wayback machine on) this was handled by a pull-up resistor
in the parallel interface, on the "off-line" signal. ANY time the printer
was powered off, set offline, or cable unplugged, the "off-line" signal
was raised by the pull-up. No data lost.
Now the parallel interface is bidirectional, and can have multiple devices
attached - this "fix" cannot be used. The interface is now more of a
buss than a single attached interface, and signals from a missing device
(powered off or disconnected) are floating. They may float high or low,
and depending on the environment (and which end of the cable is unplugged)
any thing in between.
Yes, I've lost printouts this way, but there's nothing I can really do
about it other than than tell the users "don't do that - make sure the
printer is turned on before you print".
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: [EMAIL PROTECTED]
Any opinions expressed are solely my own.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/