Re: [PATCHv5 bpf] bpf: Move iterator functions into special init section

[Jiri Olsa]

On Mon, Nov 09, 2020 at 11:04:34PM +0100, Daniel Borkmann wrote:

SNIP

> > index 7b53cb3092ee..a7c71e3b5f9a 100644
> > --- a/include/linux/init.h
> > +++ b/include/linux/init.h
> > @@ -52,6 +52,7 @@
> >   #define __initconst       __section(".init.rodata")
> >   #define __exitdata        __section(".exit.data")
> >   #define __exit_call       __used __section(".exitcall.exit")
> > +#define __init_bpf_preserve_type __section(".init.bpf.preserve_type")
> 
> Small nit, why this detour via BPF_INIT define? Couldn't we just:
> 
> #ifdef CONFIG_DEBUG_INFO_BTF
> #define __init_bpf_preserve_type   __section(".init.bpf.preserve_type")
> #else
> #define __init_bpf_preserve_type   __init
> #endif
> 
> Also, the comment above the existing defines says '/* These are for everybody 
> (although
> not all archs will actually discard it in modules) */' ... We should probably 
> not add
> the __init_bpf_preserve_type right under this listing as-is in your patch, 
> but instead
> 'separate' it by adding a small comment on top of its definition by 
> explaining its
> purpose more clearly for others.

ok, for some reason I thought I needed to add it to init.h,
but as it's bpf specific, perhaps we can omit init.h change
completely.. how about the change below?

thanks,
jirka


---
diff --git a/include/asm-generic/vmlinux.lds.h 
b/include/asm-generic/vmlinux.lds.h
index b2b3d81b1535..f91029b3443b 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -685,8 +685,21 @@
        .BTF_ids : AT(ADDR(.BTF_ids) - LOAD_OFFSET) {                   \
                *(.BTF_ids)                                             \
        }
+
+/*
+ * .init.bpf.preserve_type
+ *
+ * This section store special BPF function and marks them
+ * with begin/end symbols pair for the sake of pahole tool.
+ */
+#define INIT_BPF_PRESERVE_TYPE                                         \
+       __init_bpf_preserve_type_begin = .;                             \
+       *(.init.bpf.preserve_type)                                      \
+       __init_bpf_preserve_type_end = .;                               \
+       MEM_DISCARD(init.bpf.preserve_type)
 #else
 #define BTF
+#define INIT_BPF_PRESERVE_TYPE
 #endif
 
 /*
@@ -741,7 +754,8 @@
 #define INIT_TEXT                                                      \
        *(.init.text .init.text.*)                                      \
        *(.text.startup)                                                \
-       MEM_DISCARD(init.text*)
+       MEM_DISCARD(init.text*)                                         \
+       INIT_BPF_PRESERVE_TYPE
 
 #define EXIT_DATA                                                      \
        *(.exit.data .exit.data.*)                                      \
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 2b16bf48aab6..1739a92516ed 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -1276,10 +1276,20 @@ struct bpf_link *bpf_link_get_from_fd(u32 ufd);
 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
 int bpf_obj_get_user(const char __user *pathname, int flags);
 
+/* In case we generate BTF data, we need to group all iterator
+ * functions into special init section, so pahole can track them.
+ * Otherwise pure __init section is enough.
+ */
+#ifdef CONFIG_DEBUG_INFO_BTF
+#define __init_bpf_preserve_type __section(".init.bpf.preserve_type")
+#else
+#define __init_bpf_preserve_type __init
+#endif
+
 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
 #define DEFINE_BPF_ITER_FUNC(target, args...)                  \
        extern int bpf_iter_ ## target(args);                   \
-       int __init bpf_iter_ ## target(args) { return 0; }
+       int __init_bpf_preserve_type bpf_iter_ ## target(args) { return 0; }
 
 struct bpf_iter_aux_info {
        struct bpf_map *map;