Re: Problem with ioctl command TCGETS

From: Jan Engelhardt <jengelh@linux01.gwdg.de>
To: Miklos Szeredi <miklos@szeredi.hu>
Cc: viro@parcelfarce.linux.theplanet.co.uk,
	oebilgen@uekae.tubitak.gov.tr, linux-kernel@vger.kernel.org
Subject: Re: Problem with ioctl command TCGETS
Date: Sun, 28 Nov 2004 11:08:38 +0100 (MET)	[thread overview]
Message-ID: <Pine.LNX.4.53.0411281104100.20168@yvahk01.tjqt.qr> (raw)
In-Reply-To: <E1CYLpf-0001VQ-00@dorka.pomaz.szeredi.hu>

>On the subject of ioctls:  what about a replacement syscall:
>
> /**
>  * Getting and setting file parameters safely (ioctl done right)
>  *
>  * @fd     file descriptor
>  * @param  name of the parameter to get/set
>  * @dir    direction flag indicating either get, set, or set-get
>  * @value  value to set parameter to (set) or store current value into (get)
>  * @size   size of value
>  */
> int fparam(int fd, const char *param, int dir, void *value, size_t size);
>
>I know it's been talked about in the past.  Is anyone interested?

The idea is nice, yet the "dir" and "size" parameters in the original _IO*
macros are only there (IMO) to protect against using the wrong value for the
wrong operation on the wrong fd/file/device/socket/etc.

What is the point in making "param" a char*? You would need to parse it down
again.

Oh yeah and call it ioctl2(), sounds more cryptical :)

Jan Engelhardt
-- 
Gesellschaft für Wissenschaftliche Datenverarbeitung
Am Fassberg, 37077 Göttingen, www.gwdg.de