Re: CVS commit: syssrc/sys/kern

[Simon Burge <simonb%wasabisystems.com@localhost>]

matthew green wrote:

>    
>    > Module Name:     syssrc
>    > Committed By:    simonb
>    > Date:            Fri Nov  1 15:20:04 UTC 2002
>    > 
>    > Modified Files:
>    >  syssrc/sys/kern: subr_disk.c
>    > 
>    > Log Message:
>    > When calculating the space needed for the data, use the supplied
>    > userland structure size (if passed in).
>    > Use the supplied userland structure size (if passed in) to check if
>    > there is enough room to copyout the next structure.
>    
>    When the mib[3] wasn't specified, we should assume old struct
>    disk_sysctl instead of the current one to keep binary compatibility,
>    shouldn't we?
> 
> 
> that's a great idea!

The way I had intended this sysctls to work was if the structure size
mib entry wasn't passed in then we use the current kernel size and that
this mib entry was optional.  I guess this really only matters for
source-code compatibility with older programs that aren't written with
forward compatibility in mind.

This leaves us with a choice - do we make the size mib entry mandatory
now?  If so, is there any third party source code that would break
because it doesn't use the third entry and would fail now because all of
a sudden if doesn't get the full structure back (ucd-snmp is the only
thing that comes to mind, and it may not use the hw.diskstats sysctl
anyway).

Matt - did you say there was something else affecting binary
compatibility?  Or was this the only issue?

I just want to make sure we can everything covered if we do this.

Simon.
--
Simon Burge                            <simonb%wasabisystems.com@localhost>
NetBSD Support and Service:         http://www.wasabisystems.com/