I'm writing a Linux kernel module which needs to call a static function in the kernel. Unfortunately, the static functions are compiled with arguments passed in registers. I don't want to change the kernel, so I have to use some trick to pass arguments in different ways. Currently, I use inline assembly, but I thought maybe GCC provides a better way. By the way, how does Linux kernel handle it? As I know, static functions pass their arguments in registers, and others in the stack. How does Linux kernel use them together? Zheng Da On Jan 2, 2008 11:45 PM, Tom St Denis <tstdenis@xxxxxxxxxxxxxxxx> wrote: > Zheng Da wrote: > > Suppose I have two libraries, A and B. > > In library A, there is function a() whose arguments are passed in registers. > > In library B, there is b() whose arguments are passed in the stack. > > Now I have to use a() and b() in my program. > > But how do I tell GCC that they use different ways to pass the > > arguments and which is the right way. > > > This is why a common ABI exists. The only "sane" way to accomplish this > with GCC is to write a "thunking" [iirc that's the term] layer in which > you pass the parameters in a different convention. > > How did you end up with this mess anyways? Most assembler routines > that interface with C are written to use the ABI so they're drop in. > > Tom >
