Re: [PATCH v1] tools: fix usage of strncpy

[Ian Jackson <ian.jackson@xxxxxxxxxx>]

Olaf Hering writes ("Re: [PATCH v1] tools: fix usage of strncpy"):
> Am Mon,  8 Jun 2020 09:28:54 +0200
> schrieb Olaf Hering <olaf@xxxxxxxxx>:
> > off-by-one error in libxl__prepare_sockaddr_un
> 
> There is none, I had read the code backwards...

I have just had the same thoughts but in the opposite order.  That is
at first I thought this was not a problem, but now I think there is.

There are some kernel interfaces where a fixed-size buffer is
provided, and the kernel will tolerate a null-terminated string, but
will in any case not read beyond the end of the buffer.  Anything
involving IFNAMSIZ comes to mind.

But I think sun_path is not one of those.  The manpage I have here
says that to be portable you must null-terminate sun_path.  I know
that there are some implementations where it is possible to pass a
longer path, effectively treating sun_path as a trailing vla.

Looking at your diff, its effect seems to be to ensure
null-termination by truncating overlong paths.

I think the right approach is to return an error, not to silently
truncate.

Ian.