[PATCH 01/10] usb-serial: URB write locking macros.

[PATCH 01/10] usb-serial: URB write locking macros.

[Luiz Fernando Capitulino]

 Introduces URB write locking macros.

Signed-off-by: Luiz Capitulino <lcapitulino@mandriva.com.br>

 drivers/usb/serial/usb-serial.h |   37 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 36 insertions(+), 1 deletion(-)

diff -Nparu -X /home/lcapitulino/kernels/dontdiff a/drivers/usb/serial/usb-serial.h a~/drivers/usb/serial/usb-serial.h
--- a/drivers/usb/serial/usb-serial.h	2005-12-04 20:05:26.000000000 -0200
+++ a~/drivers/usb/serial/usb-serial.h	2005-12-04 20:06:53.000000000 -0200
@@ -17,6 +17,7 @@
 #include <linux/config.h>
 #include <linux/kref.h>
 #include <asm/semaphore.h>
+#include <asm/atomic.h>
 
 #define SERIAL_TTY_MAJOR	188	/* Nice legal number now */
 #define SERIAL_TTY_MINORS	255	/* loads of devices :) */
@@ -83,7 +84,7 @@ struct usb_serial_port {
 	unsigned char *		bulk_out_buffer;
 	int			bulk_out_size;
 	struct urb *		write_urb;
-	int			write_urb_busy;
+	atomic_t		write_urb_busy;
 	__u8			bulk_out_endpointAddress;
 
 	wait_queue_head_t	write_wait;
@@ -104,6 +105,40 @@ static inline void usb_set_serial_port_d
 	dev_set_drvdata(&port->dev, data);
 }
 
+/*
+ * usb_serial URB write access locking functions
+ *
+ * Protects 'write_urb_busy' member access, used to avoid two or more threads
+ * trying to write the same URB at the same time.
+ */
+
+/* Initialize the lock */
+static inline void usb_serial_write_urb_lock_init(struct usb_serial_port *port)
+{
+	atomic_set(&port->write_urb_busy, 0);
+}
+
+/*
+ * Lock function: returns zero if the lock was acquired,
+ * and non-zero otherwise
+ */
+static inline int usb_serial_write_urb_lock(struct usb_serial_port *port)
+{
+	return !atomic_add_unless(&port->write_urb_busy, 1, 1);
+}
+
+/* unlock function */
+static inline void usb_serial_write_urb_unlock(struct usb_serial_port *port)
+{
+	atomic_set(&port->write_urb_busy, 0);
+}
+
+/* Lock test: returns non-zero if the port is locked, and zero otherwise */
+static inline int usb_serial_write_urb_locked(struct usb_serial_port *port)
+{
+	return atomic_read(&port->write_urb_busy);
+}
+
 /**
  * usb_serial - structure used by the usb-serial core for a device
  * @dev: pointer to the struct usb_device for this device


-- 
Luiz Fernando N. Capitulino

Re: [PATCH 01/10] usb-serial: URB write locking macros.

[Arjan van de Ven]

On Tue, 2005-12-06 at 09:57 -0200, Luiz Fernando Capitulino wrote:
>  Introduces URB write locking macros.

ugh.. WHY ?

Re: [linux-usb-devel] [PATCH 01/10] usb-serial: URB write locking macros.

[Oliver Neukum]

Am Dienstag, 6. Dezember 2005 12:57 schrieb Luiz Fernando Capitulino:
> 
>  Introduces URB write locking macros.

Bluntly put, what for? What good do these changes do?

	Regards
		Oliver

Re: [PATCH 01/10] usb-serial: URB write locking macros.

[Luiz Fernando Capitulino]

 Hi Arjan,

On Tue, 06 Dec 2005 13:24:03 +0100
Arjan van de Ven <arjan@infradead.org> wrote:

| On Tue, 2005-12-06 at 09:57 -0200, Luiz Fernando Capitulino wrote:
| >  Introduces URB write locking macros.
| 
| ugh.. WHY ?

 Because is easier to read/understand:

	if (usb_serial_write_urb_lock(port)) {
		dbg("%s - already writing", __FUNCTION__);
		return 0;
	}

 than:

	if (!atomic_add_unless(&port->write_urb_busy, 1, 1)) {
		dbg("%s - already writing", __FUNCTION__);
		return 0;
	}

 IMHO.

 Of course I'm only ilustrating the 'lock' scenario, but you will have other
atomic functions spread in the driver.

 Looks better to have some macros to make clear what you're doing.

-- 
Luiz Fernando N. Capitulino

Re: [PATCH 01/10] usb-serial: URB write locking macros.

[Eduardo Pereira Habkost]

[-- Attachment #1: Type: text/plain, Size: 701 bytes --]

On Tue, Dec 06, 2005 at 01:24:03PM +0100, Arjan van de Ven wrote:
> On Tue, 2005-12-06 at 09:57 -0200, Luiz Fernando Capitulino wrote:
> >  Introduces URB write locking macros.
> 
> ugh.. WHY ?

Maybe we need a better description for the patch: the locking is already
there (the 'lock' field in struct sub_serial_port), and there is no
change of behaviour, just replacing a (spin_lock_t, int) by an atomic_t.

The purpose of the changes is removing the spinlock, because it is used
only to protect write_urb_busy, and if someday we need locking on other
parts, we already have a semaphore (introduced by Capitulino some time
ago, to fix a open()/close() race condition).

-- 
Eduardo

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]