Messages in this thread | | | Date | Wed, 7 May 2014 17:29:41 +0100 | From | One Thousand Gnomes <> | Subject | Re: [PATCH 1/2] platform: x86: dell-smo8800: Dell Latitude freefall driver (ACPI SMO8800/SMO8810) |
| |
> +static irqreturn_t smo8800_interrupt_quick(int irq, void *data) > +{ > + struct smo8800_device *smo8800 = data; > + atomic_inc(&smo8800->count); > + wake_up_interruptible(&smo8800->misc_wait); > + return IRQ_WAKE_THREAD; > +} > + > +static irqreturn_t smo8800_interrupt_thread(int irq, void *data) > +{ > + struct smo8800_device *smo8800 = data; > + dev_info(smo8800->dev, "detected free fall\n");
printk should be fast enough not to justify a thread, in fact the threaded IRQ overhead is going to be higher than the printk IMHO.
I'm not entirely sure a printk is the useful response here either ?
> +static ssize_t smo8800_misc_read(struct file *file, char __user *buf, > + size_t count, loff_t *pos) > +{ > + struct smo8800_device *smo8800 = container_of(file->private_data, > + struct smo8800_device, miscdev); > + > + DECLARE_WAITQUEUE(wait, current); > + u32 data; > + unsigned char byte_data; > + ssize_t retval = 1; > + > + if (count < 1) > + return -EINVAL;
How can this occur ??
> + > + add_wait_queue(&smo8800->misc_wait, &wait); > + while (true) { > + set_current_state(TASK_INTERRUPTIBLE); > + data = atomic_xchg(&smo8800->count, 0); > + if (data) > + break; > + > + if (file->f_flags & O_NONBLOCK) { > + retval = -EAGAIN; > + goto out; > + } > + > + if (signal_pending(current)) { > + retval = -ERESTARTSYS; > + goto out; > + } > + > + schedule(); > + }
wait_event_interruptible ?
> + if (copy_to_user(buf, &byte_data, sizeof(byte_data))) > + retval = -EFAULT;
put_user
> +static int smo8800_add(struct acpi_device *device) > +{ > + int err; > + struct smo8800_device *smo8800; > + > + if (!device) > + return -EINVAL;
How can this occur ??
> + atomic_set(&smo8800->count, 0);
Not needed - you can't see a count until it is open
> + dev_info(&device->dev, "device /dev/freefall registered with IRQ %d\n", > + smo8800->irq);
dev_dbg would be more appropriate. If every driver reported its registration we'd drown in logs
> + dev_info(&device->dev, "device /dev/freefall unregistered\n");
Ditto
Alan
| |