Re: generic strncpy - off-by-one error

[Bernd Petrovitsch <bernd@firmix.at>]

On Tue, 2003-08-12 at 19:14, Alan Cox wrote:
> On Maw, 2003-08-12 at 02:56, Anthony Truong wrote:
> > Thanks for pointing that out to me.  However, I don't think the kernel
> > strncpy implementation is exactly the same as that of Standard C lib
> 
> It is true it doesnt need to be

Nevertheless it is defined the inefficient way by the C-Standard and
lots of people actually really know this.
So IMHO (re)implementing a function called "strncpy" differently doesn't
make much sense.

> > implementation.  Iwas just looking at it from the kernel code context. 

The kernel is implemented in C. And the function strncpy() _is_ already
defined by C. So please use at least another name. Given the existence
of strlcpy() -  http://www.courtesan.com/todd/papers/strlcpy.html- this
problem already has been solved.

> > There's a point in doing it the "kernel" way, to save precious CPU
> > cycles from being wasted otherwise.
> 
> CPU cycles, got lots of those 8). If its going to do anything it might
> be to reference an extra cache line. For people who dont need padding
> 2.6 has strlcpy. Lots of drivers assume strncpy fills the entire block

ACK. To use strlcpy() is correct solution (if padding is not required).

	Bernd