Re: [PATCH 2/2] Input: Add EVIOC mechanism for MT slots

[Dmitry Torokhov]

On Fri, Jan 13, 2012 at 09:34:12AM +0100, Chase Douglas wrote:
> On 01/12/2012 08:24 PM, Henrik Rydberg wrote:
> > This patch adds the ability to extract MT slot data via a new ioctl,
> > EVIOCGMTSLOTS. The function returns an array of slot values for the
> > specified ABS_MT event type.
> > 
> > Example of user space usage:
> > 
> > struct { struct input_mt_request req; int values[64]; } data;
> > data.req.code = ABS_MT_POSITION_X;
> > if (ioctl(fd, EVIOCGMTSLOTS(sizeof(data)), &data) < 0)
> > 	return -1;
> > for (i = 0; i < data.req.num_values; i++)
> > 	printf("slot %d: %d\n", i, data.values[i]);
> 
> If input_mt_request is:
> 
> struct input_mt_request {
> 	__u32 code;
> 	__u32 num_values;
> 	__s32 values[];
> };
> 
> then how passing in the data struct above work? The passed in data would be:
> 
> __u32 code;
> __u32 num_values; /* BTW, this isn't set in the example above */
> __s32 *values; /* This is left uninitialized in the example above */
> int values[0]
> int values[1]
> int values[2]
> ...
> int values[63]
> 
> I'm not too familiar with ioctl interfaces, but I don't see how this
> example could work.

There isn't a separate __s32 *values pointer, the actual memory layout
from kernel's POV would be:

__u32 code;
__u32 num_values;
__s32 values[0];
__s32 values[1];
...
__s32 values[63];

> 
> > 
> > Cc: Chase Douglas <chase.douglas@xxxxxxxxxxxxx>
> > Signed-off-by: Henrik Rydberg <rydberg@xxxxxxxxxxx>
> > ---
> > Hi Dmitry,
> > 
> > This is a new take on the MT slot ioctl patch. A new ioctl is added,
> > which takes a request structure and returns a (variable) array of slot
> > values. The patch seems to work, but testing has not been extensive (yet).
> > 
> > Cheers,
> > Henrik
> > 
> >  drivers/input/evdev.c |   30 +++++++++++++++++++++++++++++-
> >  include/linux/input.h |   25 +++++++++++++++++++++++++
> >  2 files changed, 54 insertions(+), 1 deletions(-)
> > 
> > diff --git a/drivers/input/evdev.c b/drivers/input/evdev.c
> > index 4cf2534..8bf45d4 100644
> > --- a/drivers/input/evdev.c
> > +++ b/drivers/input/evdev.c
> > @@ -20,7 +20,7 @@
> >  #include <linux/slab.h>
> >  #include <linux/module.h>
> >  #include <linux/init.h>
> > -#include <linux/input.h>
> > +#include <linux/input/mt.h>
> >  #include <linux/major.h>
> >  #include <linux/device.h>
> >  #include "input-compat.h"
> > @@ -621,6 +621,31 @@ static int evdev_handle_set_keycode_v2(struct input_dev *dev, void __user *p)
> >  	return input_set_keycode(dev, &ke);
> >  }
> >  
> > +static int evdev_handle_mt_request(struct input_dev *dev,
> > +				   unsigned int size,
> > +				   int __user *ip)
> > +{
> > +	const struct input_mt_slot *mt = dev->mt;
> > +	unsigned int code;
> > +	int i;
> > +
> > +	if (get_user(code, &ip[0]))
> > +		return -EFAULT;
> > +	if (!input_is_mt_value(code))
> > +		return -EINVAL;
> > +
> > +	if (size < sizeof(struct input_mt_request) + dev->mtsize * sizeof(int))
> > +		return -EINVAL;
> 
> Why not fill in up to the size of the array provided?

Yes, I think we do the same with other ioctls as well.

> 
> > +	if (put_user(dev->mtsize, &ip[1]))
> > +		return -EFAULT;
> > +	for (i = 0; i < dev->mtsize; i++)
> > +		if (put_user(input_mt_get_value(&mt[i], code), &ip[2 + i]))
> > +			return -EFAULT;
> 
> This would be easier to understand if you did:
> 
> struct input_mt_request *req = (struct input_mt_request *)ip;
> 
> Then modify the stuff in a straightforward way.

This is a user pointer, you can't really do it in straightforward way...

> 
> > +	return 0;
> > +}
> > +
> >  static long evdev_do_ioctl(struct file *file, unsigned int cmd,
> >  			   void __user *p, int compat_mode)
> >  {
> > @@ -706,6 +731,9 @@ static long evdev_do_ioctl(struct file *file, unsigned int cmd,
> >  		return bits_to_user(dev->propbit, INPUT_PROP_MAX,
> >  				    size, p, compat_mode);
> >  
> > +	case EVIOCGMTSLOTS(0):
> > +		return evdev_handle_mt_request(dev, size, ip);
> > +
> >  	case EVIOCGKEY(0):
> >  		return bits_to_user(dev->key, KEY_MAX, size, p, compat_mode);
> >  
> > diff --git a/include/linux/input.h b/include/linux/input.h
> > index a637e78..9a73f18 100644
> > --- a/include/linux/input.h
> > +++ b/include/linux/input.h
> > @@ -99,6 +99,30 @@ struct input_keymap_entry {
> >  	__u8  scancode[32];
> >  };
> >  
> > +/**
> > + * struct input_mt_request - used by EVIOCGMTSLOTS ioctl
> > + * @code: ABS_MT code to read
> > + * @num_values: number of values written to the value array
> > + * @values: value array, one value per slot
> > + *
> > + * The structure is used to retrieve MT slot data. Before the call,
> > + * code is set to the wanted ABS_MT event type. On return, the value
> > + * array is filled with the slot values for the specified ABS_MT code,
> > + * and num_values is set to the actual number of slots.
> > + *
> > + * The call argument (len) is the size of the return buffer, satisfying
> > + *
> > + * len >= sizeof(input_mt_request) + sizeof(__s32) * number_of_slots
> > + *
> > + * If the request code is not an ABS_MT value, or if len is too small,
> > + * -EINVAL is returned.
> > + */
> > +struct input_mt_request {
> > +	__u32 code;
> > +	__u32 num_values;

Hmm, might avoid this one... Userspace can do EVIOCGABS to get number of
slots and then it will know how much data EVIOCGMTSLOTS will return...

> > +	__s32 values[];
> > +};
> > +
> >  #define EVIOCGVERSION		_IOR('E', 0x01, int)			/* get driver version */
> >  #define EVIOCGID		_IOR('E', 0x02, struct input_id)	/* get device ID */
> >  #define EVIOCGREP		_IOR('E', 0x03, unsigned int[2])	/* get repeat settings */
> > @@ -113,6 +137,7 @@ struct input_keymap_entry {
> >  #define EVIOCGPHYS(len)		_IOC(_IOC_READ, 'E', 0x07, len)		/* get physical location */
> >  #define EVIOCGUNIQ(len)		_IOC(_IOC_READ, 'E', 0x08, len)		/* get unique identifier */
> >  #define EVIOCGPROP(len)		_IOC(_IOC_READ, 'E', 0x09, len)		/* get device properties */
> > +#define EVIOCGMTSLOTS(len)	_IOC(_IOC_READ, 'E', 0x0a, len)		/* get MT slot values */
> >  
> >  #define EVIOCGKEY(len)		_IOC(_IOC_READ, 'E', 0x18, len)		/* get global key state */
> >  #define EVIOCGLED(len)		_IOC(_IOC_READ, 'E', 0x19, len)		/* get all LEDs */
> 

Thanks.

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