Re: [PATCH][next] ftrace: Fix -Wcast-function-type warnings on powerpc64

From: Steven Rostedt
Date: Tue Oct 05 2021 - 15:08:12 EST


On Tue, 5 Oct 2021 11:50:27 -0500
"Gustavo A. R. Silva" <gustavoars@xxxxxxxxxx> wrote:

> On Tue, Oct 05, 2021 at 12:35:22PM -0400, Steven Rostedt wrote:
> > On Tue, 5 Oct 2021 11:18:12 -0500
> > "Gustavo A. R. Silva" <gustavoars@xxxxxxxxxx> wrote:
> >
> > > On Tue, Oct 05, 2021 at 11:17:14AM -0400, Steven Rostedt wrote:
> > > > On Tue, 5 Oct 2021 00:39:22 -0500
> > > > "Gustavo A. R. Silva" <gustavoars@xxxxxxxxxx> wrote:
> > > >
> > > > > In order to make sure new function cast mismatches are not introduced
> > > > > in the kernel (to avoid tripping CFI checking), the kernel should be
> > > > > globally built with -Wcast-function-type.
> > > > >
> > > > > So, fix the following -Wcast-function-type warnings on powerpc64
> > > > > (ppc64_defconfig):
> > > >
> > > > I think I'll go back and add my linker magic.
> > > >
> > > > https://lore.kernel.org/all/20200617165616.52241bde@xxxxxxxxxxxxxxxx/
> > > >
> > > > I'll clean it up a bit too. I'll have a patch in a bit.
> > >
> > > Awesome. :)
> > >
> > > Thanks
> > > --
> >
> > Does this fix it for you?
>
> Nop; there are still some warnings (ppc64_defconfig):

Sure you applied it?

Because I tested it on powerpc64 with the defconfig, and without the patch,
I get the error, but with it, I don't.

>
> kernel/trace/ftrace.c: In function ‘ftrace_ops_get_list_func’:
> kernel/trace/ftrace.c:171:10: error: returning ‘void (*)(long unsigned int, long unsigned int, struct ftrace_ops *, struct ftrace_regs *)’ from a function with incompatible return type ‘ftrace_func_t’ {aka ‘void (*)(long unsigned int, long unsigned int)’} [-Werror=incompatible-pointer-types]
> 171 | return ftrace_ops_list_func;
> | ^~~~~~~~~~~~~~~~~~~~

Or did you not remove your patch first?

Because, the above error shows:

return type ‘ftrace_func_t’ {aka ‘void (*)(long unsigned int, long unsigned int)’

But my code has:

typedef void (*ftrace_func_t)(unsigned long ip, unsigned long parent_ip,
struct ftrace_ops *op, struct ftrace_regs *fregs);


Which is not the same type, but your patch has:

+#if ARCH_SUPPORTS_FTRACE_OPS
typedef void (*ftrace_func_t)(unsigned long ip, unsigned long parent_ip,
struct ftrace_ops *op, struct ftrace_regs *fregs);
+#else
+typedef void (*ftrace_func_t)(unsigned long ip, unsigned long parent_ip);
+#endif
+
+typedef void (*ftrace_func_base_t)(void);
+#define CAST_FTRACE_FUNC(f) ((ftrace_func_t)((ftrace_func_base_t)(f)))

Which redefines ftrace_func_t.

-- Steve


> kernel/trace/ftrace.c: In function ‘update_ftrace_function’:
> kernel/trace/ftrace.c:204:8: error: assignment to ‘ftrace_func_t’ {aka ‘void (*)(long unsigned int, long unsigned int)’} from incompatible pointer type ‘void (*)(long unsigned int, long unsigned int, struct ftrace_ops *, struct ftrace_regs *)’ [-Werror=incompatible-pointer-types]
> 204 | func = ftrace_ops_list_func;
> | ^
> kernel/trace/ftrace.c:217:11: warning: comparison of distinct pointer types lacks a cast
> 217 | if (func == ftrace_ops_list_func) {
> | ^~
> kernel/trace/ftrace.c: In function ‘ftrace_modify_all_code’:
> kernel/trace/ftrace.c:2695:35: error: passing argument 1 of ‘ftrace_update_ftrace_func’ from incompatible pointer type [-Werror=incompatible-pointer-types]
> 2695 | err = ftrace_update_ftrace_func(ftrace_ops_list_func);
> | ^~~~~~~~~~~~~~~~~~~~
> | |
> | void (*)(long unsigned int, long unsigned int, struct ftrace_ops *, struct ftrace_regs *)
> In file included from kernel/trace/ftrace.c:29:
> ./include/linux/ftrace.h:585:52: note: expected ‘ftrace_func_t’ {aka ‘void (*)(long unsigned int, long unsigned int)’} but argument is of type ‘void (*)(long unsigned int, long unsigned int, struct ftrace_ops *, struct ftrace_regs *)’
> 585 | extern int ftrace_update_ftrace_func(ftrace_func_t func);
> | ~~~~~~~~~~~~~~^~~~
> kernel/trace/ftrace.c:2705:38: warning: comparison of distinct pointer types lacks a cast
> 2705 | if (update && ftrace_trace_function != ftrace_ops_list_func) {
> | ^~
>
> --
> Gustavo