Actually I was running a kernel with only bmcsensors changed to use the void * - it should work transparently for all the other drivers (except for a compilation warning about invalid show/store pointer types being passed - and I have a perl script we can run against the source tree to update drivers and get rid of those warnings). The void * is completely optional of course, its up to the driver to use it or not (unless I'm missing your question here?). Yani On 4/28/05, Grant Coady <grant_lkml at dodo.com.au> wrote: > Hi Greg, > On Wed, 27 Apr 2005 23:57:41 -0700, Greg KH <greg at kroah.com> wrote: > > > > >And how would it work with the 99% of the kernel that doesn't create > >attributes dynamically on the fly? > > Create an index attribute: write index value, read/write attribute > data? Nothing else need know, just the driver needing unbounded > indexed attributes. > > --Grant. > >
