From: Dmitry Torokhov <dtor_core@ameritech.net>
To: Alessandro Zummo <a.zummo@towertech.it>
Subject: Re: [PATCH 5/8] RTC subsystem, dev interface
Date: Sun, 8 Jan 2006 21:50:21 -0500
User-Agent: KMail/1.8.3
Cc: linux-kernel@vger.kernel.org
References: <20060108231235.153748000@linux> <20060108231255.609424000@linux>
In-Reply-To: <20060108231255.609424000@linux>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200601082150.22213.dtor_core@ameritech.net>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 11012
Lines: 422

On Sunday 08 January 2006 18:12, Alessandro Zummo wrote:
> +
> +static int rtc_dev_open(struct inode *inode, struct file *file)
> +{
> +	int err;
> +	struct rtc_device *rtc = container_of(inode->i_cdev,
> +					struct rtc_device, char_dev);
> +	struct rtc_class_ops *ops = rtc->ops;
> +
> +	/* We keep the lock as long as the device is in use
> +	 * and return immediately if busy
> +	 */
> +	if (down_trylock(&rtc->char_sem))
> +		return -EBUSY;
> +

Does the device have to be opened for exclusively? Can it support
concurrent reads?


> +	file->private_data = &rtc->class_dev;
> +
> +	err = ops->open ? ops->open(rtc->class_dev.dev) : 0;
> +	if (err == 0) {
> +
> +		spin_lock_irq(&rtc->irq_lock);
> +		rtc->irq_data = 0;
> +		spin_unlock_irq(&rtc->irq_lock);
> +
> +		return 0;
> +	}
> +
> +	/* something has gone wrong, release the lock */
> +	up(&rtc->char_sem);
> +	return err;
> +}
> +
> +
> +static ssize_t
> +rtc_dev_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct rtc_device *rtc = to_rtc_device(file->private_data);
> +
> +	DECLARE_WAITQUEUE(wait, current);
> +	unsigned long data;
> +	ssize_t ret;
> +
> +	if (count < sizeof(unsigned long))
> +		return -EINVAL;
> +
> +	add_wait_queue(&rtc->irq_queue, &wait);
> +	do {
> +		__set_current_state(TASK_INTERRUPTIBLE);
> +
> +		spin_lock_irq(&rtc->irq_lock);
> +		data = rtc->irq_data;
> +		rtc->irq_data = 0;
> +		spin_unlock_irq(&rtc->irq_lock);
> +
> +		if (data != 0) {
> +			ret = 0;
> +			break;
> +		}
> +		if (file->f_flags & O_NONBLOCK) {
> +			ret = -EAGAIN;
> +			break;
> +		}
> +		if (signal_pending(current)) {
> +			ret = -ERESTARTSYS;
> +			break;
> +		}
> +		schedule();
> +	} while (1);
> +	set_current_state(TASK_RUNNING);
> +	remove_wait_queue(&rtc->irq_queue, &wait);
> +


The above looks very much like open-coded wait_event_interruptible();

> +	if (ret == 0) {
> +		ret = put_user(data, (unsigned long __user *)buf);
> +		if (ret == 0)
> +			ret = sizeof(unsigned long);
> +	}
> +	return ret;
> +}
> +
> +static unsigned int rtc_dev_poll(struct file *file, poll_table *wait)
> +{
> +	struct rtc_device *rtc = to_rtc_device(file->private_data);
> +	unsigned long data;
> +
> +	poll_wait(file, &rtc->irq_queue, wait);
> +
> +	spin_lock_irq(&rtc->irq_lock);
> +	data = rtc->irq_data;
> +	spin_unlock_irq(&rtc->irq_lock);
> +
> +	return data != 0 ? POLLIN | POLLRDNORM : 0;
> +}

What does the lock above protect? Once it is released rtc->irq_data may
change so reader that was woken up may not see any data anyway.

> +static int rtc_dev_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
> +		     unsigned long arg)
> +{
> +	int err = 0;
> +	struct class_device *class_dev = file->private_data;
> +	struct rtc_device *rtc = to_rtc_device(class_dev);
> +	struct rtc_class_ops *ops = rtc->ops;
> +	struct rtc_time tm;
> +	struct rtc_wkalrm alarm;
> +	void __user *uarg = (void __user *) arg;
> +
> +	/* avoid conflicting IRQ users */
> +	if (cmd == RTC_PIE_ON || cmd == RTC_PIE_OFF || cmd == RTC_IRQP_SET) {
> +		spin_lock(&rtc->irq_task_lock);
> +		if (rtc->irq_task)
> +			err = -EBUSY;
> +		spin_unlock(&rtc->irq_task_lock);
> +
> +		if (err < 0)
> +			return err;
> +	}
> +
> +	/* try the driver's ioctl interface */
> +	if (ops->ioctl) {
> +		err = ops->ioctl(class_dev->dev, cmd, arg);
> +		if (err < 0 && err != -EINVAL)
> +			return err;
> +	}
> +
> +	/* if the driver does not provide the ioctl interface
> +	 * or if that particular ioctl was not implemented
> +	 * (-EINVAL), we will try to emulate here.
> +	 */
> +
> +	switch (cmd) {
> +	case RTC_ALM_READ:
> +		if ((err = rtc_read_alarm(class_dev, &alarm)) < 0)
> +			return err;
> +
> +		if ((err = copy_to_user(uarg, &alarm.time, sizeof(tm))))
> +			return -EFAULT;
> +		break;
> +
> +	case RTC_ALM_SET:
> +		if ((err = copy_from_user(&alarm.time, uarg, sizeof(tm))))
> +			return -EFAULT;
> +
> +		alarm.enabled = 0;
> +		alarm.pending = 0;
> +		alarm.time.tm_mday = -1;
> +		alarm.time.tm_mon = -1;
> +		alarm.time.tm_year = -1;
> +		alarm.time.tm_wday = -1;
> +		alarm.time.tm_yday = -1;
> +		alarm.time.tm_isdst = -1;
> +		err = rtc_set_alarm(class_dev, &alarm);
> +		break;
> +
> +	case RTC_RD_TIME:
> +		if ((err = rtc_read_time(class_dev, &tm)) < 0)
> +			return err;
> +
> +		if ((err = copy_to_user(uarg, &tm, sizeof(tm))))
> +			return -EFAULT;
> +		break;
> +
> +	case RTC_SET_TIME:
> +		if (!capable(CAP_SYS_TIME))
> +			return -EACCES;
> +
> +		if ((err = copy_from_user(&tm, uarg, sizeof(tm))))
> +			return -EFAULT;
> +
> +		err = rtc_set_time(class_dev, &tm);
> +		break;
> +#if 0
> +	case RTC_EPOCH_SET:
> +#ifndef rtc_epoch
> +		/*
> +		 * There were no RTC clocks before 1900.
> +		 */
> +		if (arg < 1900) {
> +			err = -EINVAL;
> +			break;
> +		}
> +		if (!capable(CAP_SYS_TIME)) {
> +			err = -EACCES;
> +			break;
> +		}
> +		rtc_epoch = arg;
> +		err = 0;
> +#endif
> +		break;
> +
> +	case RTC_EPOCH_READ:
> +		err = put_user(rtc_epoch, (unsigned long __user *)uarg);
> +		break;
> +#endif
> +	case RTC_WKALM_SET:
> +		if ((err = copy_from_user(&alarm, uarg, sizeof(alarm))))
> +			return -EFAULT;
> +
> +		err = rtc_set_alarm(class_dev, &alarm);
> +		break;
> +
> +	case RTC_WKALM_RD:
> +		if ((err = rtc_read_alarm(class_dev, &alarm)) < 0)
> +			return err;
> +
> +		if ((err = copy_to_user(uarg, &alarm, sizeof(alarm))))
> +			return -EFAULT;
> +		break;
> +
> +	default:
> +		err = -EINVAL;
> +		break;
> +	}
> +
> +	return err;
> +}
> +
> +static int rtc_dev_release(struct inode *inode, struct file *file)
> +{
> +	struct rtc_device *rtc = to_rtc_device(file->private_data);
> +
> +	if (rtc->ops->release)
> +		rtc->ops->release(rtc->class_dev.dev);
> +
> +	spin_lock_irq(&rtc->irq_lock);
> +	rtc->irq_data = 0;
> +	spin_unlock_irq(&rtc->irq_lock);
> +

Why is the above needed?

> +	up(&rtc->char_sem);
> +	return 0;
> +}
> +
> +static int rtc_dev_fasync(int fd, struct file *file, int on)
> +{
> +	struct rtc_device *rtc = to_rtc_device(file->private_data);
> +	return fasync_helper(fd, file, on, &rtc->async_queue);
> +}
> +
> +static struct file_operations rtc_dev_fops = {
> +	.owner		= THIS_MODULE,
> +	.llseek		= no_llseek,
> +	.read		= rtc_dev_read,
> +	.poll		= rtc_dev_poll,
> +	.ioctl		= rtc_dev_ioctl,
> +	.open		= rtc_dev_open,
> +	.release	= rtc_dev_release,
> +	.fasync		= rtc_dev_fasync,
> +};
> +
> +static ssize_t rtc_dev_show_dev(struct class_device *class_dev, char *buf)
> +{
> +        return print_dev_t(buf, class_dev->devt);
> +}
> +static CLASS_DEVICE_ATTR(dev, S_IRUGO, rtc_dev_show_dev, NULL);
> +
> +/* insertion/removal hooks */
> +
> +static int rtc_dev_add_device(struct class_device *class_dev,
> +				struct class_interface *class_intf)
> +{
> +	struct rtc_device *rtc = to_rtc_device(class_dev);
> +
> +	if (rtc->id >= RTC_DEV_MAX) {
> +		dev_err(class_dev->dev, "too many RTCs\n");
> +		return -EINVAL;
> +	}
> +
> +	init_MUTEX(&rtc->char_sem);
> +	spin_lock_init(&rtc->irq_lock);
> +	init_waitqueue_head(&rtc->irq_queue);
> +
> +	cdev_init(&rtc->char_dev, &rtc_dev_fops);
> +	rtc->char_dev.owner = rtc->owner;
> +	class_dev->devt = MKDEV(MAJOR(rtc_devt), rtc->id);
> +
> +	if (cdev_add(&rtc->char_dev, class_dev->devt, 1)) {
> +		cdev_del(&rtc->char_dev);
> +
> +		dev_err(class_dev->dev,
> +			"failed to add char device %d:%d\n",
> +			MAJOR(class_dev->devt),
> +			MINOR(class_dev->devt));
> +
> +		class_dev->devt = MKDEV(0, 0);
> +		return -ENODEV;
> +	}
> +
> +	class_device_create_file(class_dev, &class_device_attr_dev);
> +
> +	dev_info(class_dev->dev, "rtc intf: dev (%d:%d)\n",
> +		MAJOR(class_dev->devt),
> +		MINOR(class_dev->devt));
> +
> +	kobject_hotplug(&class_dev->kobj, KOBJ_ADD);
> +

This is kobject_hotplug abuse; you are not adding a new object here.

> +	return 0;
> +}
> +
> +static void rtc_dev_remove_device(struct class_device *class_dev,
> +					struct class_interface *class_intf)
> +{
> +	struct rtc_device *rtc = to_rtc_device(class_dev);
> +
> +	class_device_remove_file(class_dev, &class_device_attr_dev);
> +
> +	if (MAJOR(class_dev->devt)) {
> +		dev_dbg(class_dev->dev, "removing char %d:%d\n",
> +			MAJOR(class_dev->devt),
> +			MINOR(class_dev->devt));
> +		cdev_del(&rtc->char_dev);
> +
> +		kobject_hotplug(&class_dev->kobj, KOBJ_REMOVE);
> +

Same here...

> +		class_dev->devt = MKDEV(0, 0);
> +	}
> +}
> +
> +/* interface registration */
> +
> +struct class_interface rtc_dev_interface = {
> +	.add = &rtc_dev_add_device,
> +	.remove = &rtc_dev_remove_device,
> +};
> +

I wonder if doing rtc dev as a class device interface is a good idea.
It may be cleaner to fold it into the core.

> +static int __init rtc_dev_init(void)
> +{
> +	int err;
> +
> +	if ((err = alloc_chrdev_region(&rtc_devt, 0, RTC_DEV_MAX, "rtc")) < 0) {
> +		printk(KERN_ERR "%s: failed to allocate char dev region\n",
> +			__FILE__);
> +		return err;
> +	}
> +
> +	if ((err = rtc_interface_register(&rtc_dev_interface)) < 0) {
> +		printk(KERN_ERR "%s: failed to register the interface\n",
> +			__FILE__);
> +		unregister_chrdev_region(rtc_devt, RTC_DEV_MAX);
> +		return err;
> +	}
> +
> +	return 0;
> +}
> +
> +static void __exit rtc_dev_exit(void)
> +{
> +	class_interface_unregister(&rtc_dev_interface);
> +
> +	unregister_chrdev_region(rtc_devt, RTC_DEV_MAX);
> +}
> +
> +module_init(rtc_dev_init);
> +module_exit(rtc_dev_exit);
> +
> +MODULE_AUTHOR("Alessandro Zummo <a.zummo@towertech.it>");
> +MODULE_DESCRIPTION("RTC class dev interface");
> +MODULE_LICENSE("GPL");
> --- linux-nslu2.orig/drivers/rtc/Kconfig	2006-01-04 01:27:14.000000000 +0100
> +++ linux-nslu2/drivers/rtc/Kconfig	2006-01-04 01:27:15.000000000 +0100
> @@ -41,6 +41,17 @@ config RTC_INTF_PROC
>  	  This driver can also be built as a module. If so, the module
>  	  will be called rtc-proc.
>  
> +config RTC_INTF_DEV
> +	tristate "dev"
> +	depends on RTC_CLASS
> +	default RTC_CLASS
> +	help
> +	  Say yes here if you want to use your RTC using the dev
> +	  interface, /dev/rtc .
> +
> +	  This driver can also be built as a module. If so, the module
> +	  will be called rtc-dev.
> +
>  comment "RTC drivers"
>  	depends on RTC_CLASS
>  
> --- linux-nslu2.orig/drivers/rtc/Makefile	2006-01-04 01:27:14.000000000 +0100
> +++ linux-nslu2/drivers/rtc/Makefile	2006-01-04 01:27:15.000000000 +0100
> @@ -7,3 +7,4 @@ obj-$(CONFIG_RTC_CLASS)		+= rtc-core.o
>  rtc-core-y			:= class.o interface.o
>  obj-$(CONFIG_RTC_INTF_SYSFS)	+= rtc-sysfs.o
>  obj-$(CONFIG_RTC_INTF_PROC)	+= rtc-proc.o
> +obj-$(CONFIG_RTC_INTF_DEV)	+= rtc-dev.o
> 
> --
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

-- 
Dmitry
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/