Skip to contentSkip to conversation

bpf: Fix build for disabled CONFIG_DEBUG_INFO_BTF option

PATCH Applied Jiri Olsa 2 participants 3 messages Jul 14, 2020

Subsystems (3)
lore
Share & tools
lore

Jiri Olsa

author reviewer
#0 lore raw

PATCH 1/2 bpf: Fix build for disabled CONFIG_DEBUG_INFO_BTF option

Stephen reported following linker warnings on powerpc build:

ld: warning: orphan section .BTF_ids' from kernel/trace/bpf_trace.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/bpf/btf.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/bpf/stackmap.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from net/core/filter.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/trace/bpf_trace.o' being placed in section `.BTF_ids'

It's because we generated .BTF_ids section even when CONFIG_DEBUG_INFO_BTF is not enabled. Fixing this by generating empty btf_id arrays for this case.

---
 include/linux/btf_ids.h | 9 +++++++++
 1 file changed, 9 insertions(+)
include/linux/btf_ids.h +9 -0
index fe019774f8a7..b3c73db9587c 100644
@@ -3,6 +3,8 @@
 #ifndef _LINUX_BTF_IDS_H
 #define _LINUX_BTF_IDS_H
 
+#ifdef CONFIG_DEBUG_INFO_BTF
+
 #include <linux/compiler.h> /* for __PASTE */
 
 /*
@@ -83,5 +85,12 @@ asm(							\
 ".zero 4                                       \n"	\
 ".popsection;                                  \n");
 
+#else
+
+#define BTF_ID_LIST(name) static u32 name[5];
+#define BTF_ID(prefix, name)
+#define BTF_ID_UNUSED
+
+#endif /* CONFIG_DEBUG_INFO_BTF */
 
 #endif
Signature · 1 line
2.25.4
author reviewer
#1 lore raw

PATCH 2/2 bpf: Fix cross build for CONFIG_DEBUG_INFO_BTF option

Stephen and 0-DAY CI Kernel Test Service reported broken cross build for arm (arm-linux-gnueabi-gcc (GCC) 9.3.0), with following output:

/tmp/ccMS5uth.s: Assembler messages:
/tmp/ccMS5uth.s:69: Error: unrecognized symbol type "" /tmp/ccMS5uth.s:82: Error: unrecognized symbol type ""

Having '@object' for .type diretive is wrong because '@' is comment character for some architectures. Using STT_OBJECT instead that should work everywhere.

Also using HOST* variables to build resolve_btfids so it's properly build in crossbuilds (stolen from objtool's Makefile).

---
 include/linux/btf_ids.h           |  2 +-
 tools/bpf/resolve_btfids/Makefile | 14 ++++++++++++++
 2 files changed, 15 insertions(+), 1 deletion(-)
include/linux/btf_ids.h +1 -1
index b3c73db9587c..1cdb56950ffe 100644
@@ -23,7 +23,7 @@
 asm(							\
 ".pushsection " BTF_IDS_SECTION ",\"a\";       \n"	\
 ".local " #symbol " ;                          \n"	\
-".type  " #symbol ", @object;                  \n"	\
+".type  " #symbol ", STT_OBJECT;               \n"	\
 ".size  " #symbol ", 4;                        \n"	\
 #symbol ":                                     \n"	\
 ".zero 4                                       \n"	\
tools/bpf/resolve_btfids/Makefile +14 -0
index 948378ca73d4..a88cd4426398 100644
@@ -16,6 +16,20 @@ else
   MAKEFLAGS=--no-print-directory
 endif
 
+# always use the host compiler
+ifneq ($(LLVM),)
+HOSTAR  ?= llvm-ar
+HOSTCC  ?= clang
+HOSTLD  ?= ld.lld
+else
+HOSTAR  ?= ar
+HOSTCC  ?= gcc
+HOSTLD  ?= ld
+endif
+AR       = $(HOSTAR)
+CC       = $(HOSTCC)
+LD       = $(HOSTLD)
+
 OUTPUT ?= $(srctree)/tools/bpf/resolve_btfids/
 
 LIBBPF_SRC := $(srctree)/tools/lib/bpf/
Signature · 1 line
2.25.4
#2 lore raw
In reply to Jiri Olsa
Quoted 9 lines · Jiri Olsa

Stephen reported following linker warnings on powerpc build:

ld: warning: orphan section .BTF_ids' from kernel/trace/bpf_trace.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/bpf/btf.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/bpf/stackmap.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from net/core/filter.o' being placed in section `.BTF_ids'
ld: warning: orphan section .BTF_ids' from kernel/trace/bpf_trace.o' being placed in section `.BTF_ids'

Quoted 7 lines

It's because we generated .BTF_ids section even when CONFIG_DEBUG_INFO_BTF is not enabled. Fixing this by generating empty btf_id arrays for this case.

Applied both to bpf-next. Thanks for the quick fix.