When building bpftool with !CONFIG_PERF_EVENTS:
skeleton/pid_iter.bpf.c:47:14: error: incomplete definition of type 'struct bpf_perf_link'
perf_link = container_of(link, struct bpf_perf_link, link);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:74:22: note: expanded from macro 'container_of'
((type *)(__mptr - offsetof(type, member))); \
^~~~~~~~~~~~~~~~~~~~~~
tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:68:60: note: expanded from macro 'offsetof'
#define offsetof(TYPE, MEMBER) ((unsigned long)&((TYPE *)0)->MEMBER)
~~~~~~~~~~~^
skeleton/pid_iter.bpf.c:44:9: note: forward declaration of 'struct bpf_perf_link'
struct bpf_perf_link *perf_link;
^
&bpf_perf_link is being defined and used only under the ifdef.
Move it out of the block and explicitly emit a BTF to fix
compilation.
Fixes: cbdaf71f7e65 ("bpftool: Add bpf_cookie to link output")
Signed-off-by: Alexander Lobakin <[email protected]>
---
kernel/bpf/syscall.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index e9621cfa09f2..34fdf27d14cf 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2952,12 +2952,12 @@ static const struct bpf_link_ops bpf_raw_tp_link_lops = {
.fill_link_info = bpf_raw_tp_link_fill_link_info,
};
-#ifdef CONFIG_PERF_EVENTS
struct bpf_perf_link {
struct bpf_link link;
struct file *perf_file;
};
+#ifdef CONFIG_PERF_EVENTS
static void bpf_perf_link_release(struct bpf_link *link)
{
struct bpf_perf_link *perf_link = container_of(link, struct bpf_perf_link, link);
@@ -4333,6 +4333,7 @@ static int link_create(union bpf_attr *attr, bpfptr_t uattr)
#endif
case BPF_PROG_TYPE_PERF_EVENT:
case BPF_PROG_TYPE_TRACEPOINT:
+ BTF_TYPE_EMIT(struct bpf_perf_link);
ret = bpf_perf_link_attach(attr, prog);
break;
case BPF_PROG_TYPE_KPROBE:
--
2.35.2
On Thu, Apr 14, 2022 at 3:45 PM Alexander Lobakin <[email protected]> wrote:
>
> When building bpftool with !CONFIG_PERF_EVENTS:
>
> skeleton/pid_iter.bpf.c:47:14: error: incomplete definition of type 'struct bpf_perf_link'
> perf_link = container_of(link, struct bpf_perf_link, link);
> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:74:22: note: expanded from macro 'container_of'
> ((type *)(__mptr - offsetof(type, member))); \
> ^~~~~~~~~~~~~~~~~~~~~~
> tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:68:60: note: expanded from macro 'offsetof'
> #define offsetof(TYPE, MEMBER) ((unsigned long)&((TYPE *)0)->MEMBER)
> ~~~~~~~~~~~^
> skeleton/pid_iter.bpf.c:44:9: note: forward declaration of 'struct bpf_perf_link'
> struct bpf_perf_link *perf_link;
> ^
>
> &bpf_perf_link is being defined and used only under the ifdef.
> Move it out of the block and explicitly emit a BTF to fix
> compilation.
>
> Fixes: cbdaf71f7e65 ("bpftool: Add bpf_cookie to link output")
> Signed-off-by: Alexander Lobakin <[email protected]>
Similar to v1, this fix is weird to me. I hope we have can fix it in user
space.
From: Song Liu <[email protected]>
Date: Fri, 15 Apr 2022 16:24:41 -0700
> On Thu, Apr 14, 2022 at 3:45 PM Alexander Lobakin <[email protected]> wrote:
> >
> > When building bpftool with !CONFIG_PERF_EVENTS:
> >
> > skeleton/pid_iter.bpf.c:47:14: error: incomplete definition of type 'struct bpf_perf_link'
> > perf_link = container_of(link, struct bpf_perf_link, link);
> > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:74:22: note: expanded from macro 'container_of'
> > ((type *)(__mptr - offsetof(type, member))); \
> > ^~~~~~~~~~~~~~~~~~~~~~
> > tools/bpf/bpftool/bootstrap/libbpf/include/bpf/bpf_helpers.h:68:60: note: expanded from macro 'offsetof'
> > #define offsetof(TYPE, MEMBER) ((unsigned long)&((TYPE *)0)->MEMBER)
> > ~~~~~~~~~~~^
> > skeleton/pid_iter.bpf.c:44:9: note: forward declaration of 'struct bpf_perf_link'
> > struct bpf_perf_link *perf_link;
> > ^
> >
> > &bpf_perf_link is being defined and used only under the ifdef.
> > Move it out of the block and explicitly emit a BTF to fix
> > compilation.
> >
> > Fixes: cbdaf71f7e65 ("bpftool: Add bpf_cookie to link output")
> > Signed-off-by: Alexander Lobakin <[email protected]>
>
> Similar to v1, this fix is weird to me. I hope we have can fix it in user
> space.
I've been thinking on this, but userspace is not provided with any
autoconf.h definitions (only selftests have them), so its code must
be sort of universal.
Both this and 01/11 are compile time and due to imcomplete and/or
absent BTF struct declarations. I'm not familiar with
bpf_core_field_exists(), and it might be that it's able to solve
01/11, but not this one. &bpf_perf_link must be present
unconditionally, otherwise it won't be defined in the generated
vmlinux.h at all.
Thanks,
Al