failing i2c driver on mpc8xx - interruptible_sleep_on()

Hello all,

I'm attempting to write a driver to communicate on the i2c bus with a single clock chip. The mpc8xx is the master device and the clock chip is the slave. Very simple. The clock chip is a ds1307 from Dallas Semiconductor. Linux kernel version is 2.4.4, modified by Denx Software Engineering.

I'm using functions in i2c-core.c and i2c-algo-8xx.c. To write a message (pointed to by msg) to the i2c bus, I call:

i2c_transfer(ds1307_client->adapter, msg, 1);

which is to be found in i2c-core.c and does the following:

       I2C_LOCK(adap);

       ret = adap->algo->master_xfer(adap,msgs,num);

       I2C_UNLOCK(adap);

This locks the i2c adapter (which is a conceptual entity) and then calls master_xfer(), which is a pointer to cpm_xfer(), which you can find in i2c-algo-8xx.c. cpm_xfer() sorts some variables and such like and then calls:

cpm_iic_write(), from i2c-algo-8xx.c

This function contains a line:

       interruptible_sleep_on(&iic_wait);

Which is causing my mpc8xx to hang (no serial console response, no telnet access).

cpm_iic_read(), also has an interruptible_sleep_on() call which hangs my target in the same way.

Does anyone know why this might be? interruptible_sleep_on() is in sched.c and looks like:

void interruptible_sleep_on(wait_queue_head_t *q)

{
	SLEEP_ON_VAR

       // SEB: Makes the task an interuptible task, so another task can preempt it
	current->state = TASK_INTERRUPTIBLE;

	SLEEP_ON_HEAD
	schedule();
	SLEEP_ON_TAIL
}

where
SLEEP_ON_VAR
SLEEP_ON_HEAD
SLEEP_ON_TAIL
expand to:

#define	SLEEP_ON_VAR				\
	unsigned long flags;			\
	wait_queue_t wait;			\
	init_waitqueue_entry(&wait, current);

#define	SLEEP_ON_HEAD					\
	wq_write_lock_irqsave(&q->lock,flags);		\
	__add_wait_queue(q, &wait);			\
	wq_write_unlock(&q->lock);

#define	SLEEP_ON_TAIL						\
	wq_write_lock_irq(&q->lock);				\
	__remove_wait_queue(q, &wait);				\
	wq_write_unlock_irqrestore(&q->lock,flags);

I am stepping through the code using a bdm debugger and gdb. The debugger just waits when I try to step into interriptible_sleep_on().

Any help that anyone could offer such as suggestions on how to work out what this problem is about will be gratefully received.

thanks,

Seb James.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Seb,

What's going on is that when you call interruptible_sleep_on(), you're putting the process to sleep until an interrupt comes in and wakes it up. iic_wait is the wait queue that you're using. So in order for the process to continue, and interrupt has to occur and wake this process up.

The most likely reason that you're hanging is because the interrupt is never coming in.

--Chris

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/