Re: ioctl cleanups: enable sg_io and serial stuff to be shared

Pavel Machek (pavel@ucw.cz)
Thu, 8 May 2003 22:33:14 +0200


Hi!

> > Not sure if we are not too close to stable release to do that? And I
> > see no incremental way how to get there. Moving compatibility stuff
> > closer to drivers can be done close to stable release...
>
> You can define it as follows:
>
> 1) If entry exists in COMPAT or TRANSLATE table, invoke
> fops->ioctl(), else
>
> 2) If ->compat_ioctl() exists, invoke that, else
>
> 3) Fail.
>
> The COMPAT tables are sort of valuable, in that it eliminates
> the need to duplicate code when all of a drivers ioctls need
> no translation.
>
> BTW, need to extend this to netdev->do_ioctl as well for the
> handling of SIOCDEVPRIVATE based stuff. Oh goody, we can finally
> fix up that crap :))))

There's a *lot* of structs that contain *ioctl:
pavel@amd:/usr/src/linux-test/include/linux$ grep "*ioctl" *
pavel@amd:/usr/src/linux-test/include/linux$ grep "*ioctl" *
atmdev.h: int (*ioctl)(struct atm_dev *dev,unsigned int cmd,void *arg);
atmdev.h: int (*ioctl)(struct atm_dev *dev,unsigned int cmd,void *arg);
fs.h: int (*ioctl) (struct inode *, struct file *, unsigned, unsigned long);
fs.h: int (*ioctl) (struct inode *, struct file *, unsigned int, unsigned long);
hdlc.h: int (*ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
hdlcdrv.h: int (*ioctl)(struct net_device *, struct ifreq *,
ide.h: int (*ioctl)(ide_drive_t *, struct inode *, struct file *, unsigned int, unsigned long);
if_bridge.h:extern void brioctl_set(int (*ioctl_hook)(unsigned long));
if_pppox.h: int (*ioctl)(struct socket *sock, unsigned int cmd,
ioctl32.h:typedef int (*ioctl_trans_handler_t)(unsigned int, unsigned int, unsigned long, struct file *);
loop.h: int (*ioctl)(struct loop_device *, int cmd,
loop.h: int (*ioctl)(struct loop_device *, int cmd, unsigned long arg);
net.h: int (*ioctl) (struct socket *sock, unsigned int cmd,
ppp_channel.h: int (*ioctl)(struct ppp_channel *, unsigned int, unsigned long);
serial_core.h: int (*ioctl)(struct uart_port *, unsigned int, unsigned long);
tty_driver.h: * int (*ioctl)(struct tty_struct *tty, struct file * file,
tty_driver.h: int (*ioctl)(struct tty_struct *tty, struct file * file,
tty_ldisc.h: * int (*ioctl)(struct tty_struct * tty, struct file * file,
tty_ldisc.h: int (*ioctl)(struct tty_struct * tty, struct file * file,
usb.h: int (*ioctl) (struct usb_interface *intf, unsigned int code, void *buf);
wanrouter.h: int (*ioctl) (struct wan_device *wandev, unsigned cmd,
pavel@amd:/usr/src/linux-test/include/linux$

What about this one: redefine it to (*ioctl)( ...., unsigned *long*,
unsinged long). That means we can add

#define IOCTL_COMPAT 0x1 0000 0000

and avoid adding new field to each such structure. Also I will not
have to duplicate lots of middle-level code (I will have to modify
unsigned int -> unsigned long, but no second copies of everything). It
means that architecture with CONFIG_COMPAT needs to have unsigned long
> 32 bits, but I guess we can live with that.

What do you think?
Pavel

-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]
-
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/