| Message ID | 20201109185754.377373-1-jolsa@kernel.org |
|---|---|
| State | New |
| Headers | show |
| Series | [PATCHv5,bpf] bpf: Move iterator functions into special init section | expand |
On 11/9/20 7:57 PM, Jiri Olsa wrote: > With upcoming changes to pahole, that change the way how and > which kernel functions are stored in BTF data, we need a way > to recognize iterator functions. > > Iterator functions need to be in BTF data, but have no real > body and are currently placed in .init.text section, so they > are freed after kernel init and are filtered out of BTF data > because of that. > > The solution is to place these functions under new section: > .init.bpf.preserve_type > > And add 2 new symbols to mark that area: > __init_bpf_preserve_type_begin > __init_bpf_preserve_type_end > > The code in pahole responsible for picking up the functions will > be able to recognize functions from this section and add them to > the BTF data and filter out all other .init.text functions. > > Cc: Arnaldo Carvalho de Melo <acme@kernel.org> > Suggested-by: Yonghong Song <yhs@fb.com> > Acked-by: Song Liu <songliubraving@fb.com> > Acked-by: Andrii Nakryiko <andrii@kernel.org> > Signed-off-by: Jiri Olsa <jolsa@redhat.com> > --- > v5 changes: > - use "" in __section macro due to: > 33def8498fdde180 ("treewide: Convert macro and uses of __section(foo) to __section("foo")") > [Arnaldo] > > v4: https://lore.kernel.org/bpf/20201106222512.52454-1-jolsa@kernel.org/ > > include/asm-generic/vmlinux.lds.h | 16 +++++++++++++++- > include/linux/bpf.h | 8 +++++++- > include/linux/init.h | 1 + > 3 files changed, 23 insertions(+), 2 deletions(-) > > 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..73e8ededde3e 100644 > --- a/include/linux/bpf.h > +++ b/include/linux/bpf.h > @@ -1276,10 +1276,16 @@ 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); > > +#ifdef CONFIG_DEBUG_INFO_BTF > +#define BPF_INIT __init_bpf_preserve_type > +#else > +#define BPF_INIT __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 BPF_INIT bpf_iter_ ## target(args) { return 0; } > > struct bpf_iter_aux_info { > struct bpf_map *map; > diff --git a/include/linux/init.h b/include/linux/init.h > 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. > /* > * modpost check for section mismatches during the kernel build. >
On 11/10/20 11:35 AM, Jiri Olsa wrote: > 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? Agree, that looks much better, thanks!
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..73e8ededde3e 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -1276,10 +1276,16 @@ 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); +#ifdef CONFIG_DEBUG_INFO_BTF +#define BPF_INIT __init_bpf_preserve_type +#else +#define BPF_INIT __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 BPF_INIT bpf_iter_ ## target(args) { return 0; } struct bpf_iter_aux_info { struct bpf_map *map; diff --git a/include/linux/init.h b/include/linux/init.h 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") /* * modpost check for section mismatches during the kernel build.