Jason Thorpe <thorpej@nas.nasa.gov> writes:

> On 24 Jul 1998 18:04:45 -0700 
>  cgd@netbsd1.cygnus.com (Chris G. Demetriou) wrote:
> 
>  > "R vs. W don't do what you think they do."
>  > 
>  > In particular, they have _nothing_ to do with checking permissions
>  > (you have to do that yourself with the ioctl flags arg, if you think
>  > it's appropriate -- in this case, given the way keyboards are used,
>  > i'm not sure that it is), but rather with whether data is to be copied
>  > into or out of the kernel on ioctl() entry and exit.
>  > 
>  > _IO == no data
>  > _IOR == kernel reads the data on entry
>  > _IOW == kernel writes the data on exit
>  > _IOWR == kernel reads on entry then writes on exit.
> 
> not according to ioccom.h
> 
> IOC_IN == copy parameters in, and _IOW uses IOC_IN.  _IOR uses IOC_OUT.
> 
> I think the "read/write" is from the perspective of the requesting process.

*chuckle* "D'OH!  Well!  Now you know why the code was broken to begin
with..."  8-)


chris