[PATCH] Regression in 2.6.13 in USB generic serial driver

[PATCH] Regression in 2.6.13 in USB generic serial driver

[Randall Nortman]

Kernel version 2.6.13 introduced a regression (vs. 2.6.12.6) in the
generic USB serial converter driver.  The regression manifests, as far
as I can tell, whenever you attempt to write to the device.  The
symptom is that the write never completes (write() returns 0, or
blocks).  I'm pretty sure this will happen with all hardware, but I
only have my hardware to test with (amd64 host interfacing to a
custom-built USB device based on an Atmel AT91SAM7X256 arm7tdmi chip's
built-in USB transceiver).

The patch (vs. vanilla 2.6.13.4) is rather straightforward:

===================================================================
diff -ru linux-2.6.13.4/drivers/usb/serial/generic.c linux-2.6.13.4-fixed/drivers/usb/serial/generic.c
--- linux-2.6.13.4/drivers/usb/serial/generic.c	2005-10-10 14:54:29.000000000 -0400
+++ linux-2.6.13.4-fixed/drivers/usb/serial/generic.c	2005-10-13 21:34:54.000000000 -0400
@@ -223,7 +223,7 @@
 	dbg("%s - port %d", __FUNCTION__, port->number);
 
 	if (serial->num_bulk_out) {
-		if (port->write_urb_busy)
+		if (!(port->write_urb_busy))
 			room = port->bulk_out_size;
 	}
 
===================================================================

Please let me know if I'm missing something.  Otherwise, do I need to
do anything in particular to get a maintainer to accept the patch?
(I'm a long-time user, first-time patcher, so forgive my ignorance
about procedures.)

-- 
Randall

Re: [PATCH] Regression in 2.6.13 in USB generic serial driver

[Chris Wright]

* Randall Nortman (linuxkernellist@wonderclown.com) wrote:
> Please let me know if I'm missing something.  Otherwise, do I need to
> do anything in particular to get a maintainer to accept the patch?
> (I'm a long-time user, first-time patcher, so forgive my ignorance
> about procedures.)

Take a look at Documentation/SubmittingPatches in the kernel tree.
You'll find info on MAINTAINERS, patch formats, and signing off on your
changes.

thanks,
-chris

[PATCH] usbserial: Regression in USB generic serial driver

[Randall Nortman]

Kernel version 2.6.13 introduced a regression in the generic USB
serial converter driver (usbserial.o, drivers/usb/serial/generic.c).
The bug manifests, as far as I can tell, whenever you attempt to write
to the device -- the write will never complete (write() returns 0, or
blocks).

Signed-off-by: Randall Nortman <oss@wonderclown.org>
---
I'm sending this a second time, after reading up on the right way to
submit a patch (much thanks to Chris Wright).  Now we see if I'm
capable of following instructions properly.

diff --git a/drivers/usb/serial/generic.c b/drivers/usb/serial/generic.c
--- a/drivers/usb/serial/generic.c
+++ b/drivers/usb/serial/generic.c
@@ -223,7 +223,7 @@ int usb_serial_generic_write_room (struc
 	dbg("%s - port %d", __FUNCTION__, port->number);
 
 	if (serial->num_bulk_out) {
-		if (port->write_urb_busy)
+		if (!(port->write_urb_busy))
 			room = port->bulk_out_size;
 	}