Re: [SOLVED] Serial buffer corruption [was Re: FTDI usb-serial possible bug]

[Gene Heskett <gene.heskett@gmail.com>]

On Saturday 05 May 2007, Oliver Neukum wrote:
>Am Samstag, 5. Mai 2007 20:08 schrieb Antonino Ingargiola:
>> Now I don't want to abuse your kindness, but I (personally) would be
>> *really* interested in a similar fix for the FTDI usb-serial driver,
>> because many measurements I do use an FTDI device.
>
>Does this work?
>
>	Regards
>		Oliver
>----
>
>--- a/drivers/usb/serial/ftdi_sio.c	2007-05-05 20:21:41.000000000 +0200
>+++ b/drivers/usb/serial/ftdi_sio.c	2007-05-05 20:27:09.000000000 +0200
>@@ -1749,10 +1749,6 @@ static void ftdi_process_read (struct wo
> 			length = 0;
> 		}
>
>-		if (priv->rx_flags & THROTTLED) {
>-			dbg("%s - throttled", __FUNCTION__);
>-			break;
>-		}
> 		if (tty_buffer_request_room(tty, length) < length) {
> 			/* break out & wait for throttling/unthrottling to happen */
> 			dbg("%s - receive room low", __FUNCTION__);
>@@ -1825,16 +1821,6 @@ static void ftdi_process_read (struct wo
> 		dbg("%s - incomplete, %d bytes processed, %d remain",
> 				__FUNCTION__, packet_offset,
> 				urb->actual_length - packet_offset);
>-		/* check if we were throttled while processing */
>-		spin_lock_irqsave(&priv->rx_lock, flags);
>-		if (priv->rx_flags & THROTTLED) {
>-			priv->rx_flags |= ACTUALLY_THROTTLED;
>-			spin_unlock_irqrestore(&priv->rx_lock, flags);
>-			dbg("%s - deferring remainder until unthrottled",
>-					__FUNCTION__);
>-			return;
>-		}
>-		spin_unlock_irqrestore(&priv->rx_lock, flags);
> 		/* if the port is closed stop trying to read */
> 		if (port->open_count > 0){
> 			/* delay processing of remainder */
>@@ -1856,9 +1842,15 @@ static void ftdi_process_read (struct wo
> 			      port->read_urb->transfer_buffer,
> port->read_urb->transfer_buffer_length, ftdi_read_bulk_callback, port);
>
>-		result = usb_submit_urb(port->read_urb, GFP_ATOMIC);
>-		if (result)
>-			err("%s - failed resubmitting read urb, error %d", __FUNCTION__,
> result); +		spin_lock_irqsave(&priv->rx_lock, flags);
>+		if (priv->rx_flags & THROTTLED) {
>+			priv->rx_flags |= ACTUALLY_THROTTLED;
>+		} else {
>+			result = usb_submit_urb(port->read_urb, GFP_ATOMIC);
>+			if (result)
>+				err("%s - failed resubmitting read urb, error %d", __FUNCTION__,
> result); +		}
>+		spin_unlock_irqrestore(&priv->rx_lock, flags);
> 	}
>
> 	return;
>-

Oliver:

Sort of late, but,

I have a couple of kernels building with this patch applied, one with the 
sd-0.48 patch where FTDI stuff is a 40+% cpu hog and can't be used, and one 
with the cfs-v10 patch where it can be co-erced into working if I screw with 
it enough.

I have also been using another serial patch that for unk reasons, also helps 
the usb-serial case here, but vger won't let me repost that without stripping 
off the headers.  That patch has some resemblance to 
this:8250-clear-on-read-bits-LSR-MSR where I usually replace spaces with 
dashes so I don't have to use "to surround the name" in my scripts.

So these test kernels will have both patches.

I'll let you know how they work later this morning.

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Oh my GOD -- the SUN just fell into YANKEE STADIUM!!