diff mbox series

[RFC] Revert "bpf: Fix ORC unwinding in non-JIT BPF code"

Message ID 20190708124547.3515538-1-arnd@arndb.de
State New
Headers show
Series [RFC] Revert "bpf: Fix ORC unwinding in non-JIT BPF code" | expand

Commit Message

Arnd Bergmann July 8, 2019, 12:45 p.m. UTC
Apparently this was a bit premature, at least I still get this
warning with gcc-8.1:

kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x44d2: sibling call from callable instruction with modified stack frame

This reverts commit b22cf36c189f31883ad0238a69ccf82aa1f3b16b.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>

---
 kernel/bpf/core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

-- 
2.20.0

Comments

Josh Poimboeuf July 8, 2019, 1:11 p.m. UTC | #1
On Mon, Jul 08, 2019 at 02:45:23PM +0200, Arnd Bergmann wrote:
> Apparently this was a bit premature, at least I still get this

> warning with gcc-8.1:

> 

> kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x44d2: sibling call from callable instruction with modified stack frame

> 

> This reverts commit b22cf36c189f31883ad0238a69ccf82aa1f3b16b.

> 

> Signed-off-by: Arnd Bergmann <arnd@arndb.de>


Yes, I have been working on a fix.

The impact is that ORC unwinding is broken in this function for
CONFIG_RETPOLINE=n.

I don't think we want to revert this patch though, because that will
broaden the impact to the CONFIG_RETPOLINE=y case.  Anyway I hope to
have fixes soon.

-- 
Josh
Arnd Bergmann July 8, 2019, 1:17 p.m. UTC | #2
On Mon, Jul 8, 2019 at 3:11 PM Josh Poimboeuf <jpoimboe@redhat.com> wrote:
>

> On Mon, Jul 08, 2019 at 02:45:23PM +0200, Arnd Bergmann wrote:

> > Apparently this was a bit premature, at least I still get this

> > warning with gcc-8.1:

> >

> > kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x44d2: sibling call from callable instruction with modified stack frame

> >

> > This reverts commit b22cf36c189f31883ad0238a69ccf82aa1f3b16b.

> >

> > Signed-off-by: Arnd Bergmann <arnd@arndb.de>

>

> Yes, I have been working on a fix.

>

> The impact is that ORC unwinding is broken in this function for

> CONFIG_RETPOLINE=n.

>

> I don't think we want to revert this patch though, because that will

> broaden the impact to the CONFIG_RETPOLINE=y case.  Anyway I hope to

> have fixes soon.


Ok, sounds good. Thanks,

     Arnd
diff mbox series

Patch

diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 7e98f36a14e2..16079550db6d 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -1299,7 +1299,7 @@  static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn, u64 *stack)
 {
 #define BPF_INSN_2_LBL(x, y)    [BPF_##x | BPF_##y] = &&x##_##y
 #define BPF_INSN_3_LBL(x, y, z) [BPF_##x | BPF_##y | BPF_##z] = &&x##_##y##_##z
-	static const void * const jumptable[256] __annotate_jump_table = {
+	static const void *jumptable[256] = {
 		[0 ... 255] = &&default_label,
 		/* Now overwrite non-defaults ... */
 		BPF_INSN_MAP(BPF_INSN_2_LBL, BPF_INSN_3_LBL),
@@ -1558,6 +1558,7 @@  static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn, u64 *stack)
 		BUG_ON(1);
 		return 0;
 }
+STACK_FRAME_NON_STANDARD(___bpf_prog_run); /* jump table */
 
 #define PROG_NAME(stack_size) __bpf_prog_run##stack_size
 #define DEFINE_BPF_PROG_RUN(stack_size) \