[PATCH bpf] tools/bpf: only set obj->skeleton without err

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



After `bpftool gen skeleton`, the ${bpf_app}.skel.h will provide that
${bpf_app_name}__open helper to load bpf. If there is some error
like ENOMEM, the ${bpf_app_name}__open will rollback(free) the allocated
object, including `bpf_object_skeleton`.

Since the ${bpf_app_name}__create_skeleton set the obj->skeleton first
and not rollback it when error, it will cause double-free in
${bpf_app_name}__destory at ${bpf_app_name}__open. Therefore, we should
set the obj->skeleton before return 0;

Signed-off-by: Wei Fu <fuweid89@xxxxxxxxx>
---
 tools/bpf/bpftool/gen.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/bpf/bpftool/gen.c b/tools/bpf/bpftool/gen.c
index 5c18351290f0..e61e08f524da 100644
--- a/tools/bpf/bpftool/gen.c
+++ b/tools/bpf/bpftool/gen.c
@@ -928,7 +928,6 @@ static int do_skeleton(int argc, char **argv)
 			s = (struct bpf_object_skeleton *)calloc(1, sizeof(*s));\n\
 			if (!s)						    \n\
 				goto err;				    \n\
-			obj->skeleton = s;				    \n\
 									    \n\
 			s->sz = sizeof(*s);				    \n\
 			s->name = \"%1$s\";				    \n\
@@ -1001,6 +1000,8 @@ static int do_skeleton(int argc, char **argv)
 									    \n\
 			s->data = (void *)%2$s__elf_bytes(&s->data_sz);	    \n\
 									    \n\
+			obj->skeleton = s;				    \n\
+									    \n\
 			return 0;					    \n\
 		err:							    \n\
 			bpf_object__destroy_skeleton(s);		    \n\
-- 
2.25.1




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux