From: Menglong Dong <imagedong@tencent.com>

I think there is something wrong with BPF_PROBE_MEM in ___bpf_prog_run()
in big-endian machine. Let's make a test and see what will happen if we
want to load a 'u16' with BPF_PROBE_MEM.

Let's make the src value '0x0001', the value of dest register will become
0x0001000000000000, as the value will be loaded to the first 2 byte of
DST with following code:

  bpf_probe_read_kernel(&DST, SIZE, (const void *)(long) (SRC + insn->off));

Obviously, the value in DST is not correct. In fact, we can compare
BPF_PROBE_MEM with LDX_MEM_H:

  DST = *(SIZE *)(unsigned long) (SRC + insn->off);

If the memory load is done by LDX_MEM_H, the value in DST will be 0x1 now.

And I think this error results in the test case 'test_bpf_sk_storage_map'
failing:

  test_bpf_sk_storage_map:PASS:bpf_iter_bpf_sk_storage_map__open_and_load 0 nsec
  test_bpf_sk_storage_map:PASS:socket 0 nsec
  test_bpf_sk_storage_map:PASS:map_update 0 nsec
  test_bpf_sk_storage_map:PASS:socket 0 nsec
  test_bpf_sk_storage_map:PASS:map_update 0 nsec
  test_bpf_sk_storage_map:PASS:socket 0 nsec
  test_bpf_sk_storage_map:PASS:map_update 0 nsec
  test_bpf_sk_storage_map:PASS:attach_iter 0 nsec
  test_bpf_sk_storage_map:PASS:create_iter 0 nsec
  test_bpf_sk_storage_map:PASS:read 0 nsec
  test_bpf_sk_storage_map:FAIL:ipv6_sk_count got 0 expected 3
  $10/26 bpf_iter/bpf_sk_storage_map:FAIL

The code of the test case is simply, it will load sk->sk_family to the
register with BPF_PROBE_MEM and check if it is AF_INET6. With this patch,
now the test case 'bpf_iter' can pass:

  $10  bpf_iter:OK

Fixes: 2a02759ef5f8 ("bpf: Add support for BTF pointers to interpreter")
Reviewed-by: Jiang Biao <benbjiang@tencent.com>
Reviewed-by: Hao Peng <flyingpeng@tencent.com>
Signed-off-by: Menglong Dong <imagedong@tencent.com>
---
v2:
- fold the code into LDST (Daniel Borkmann)
- add the 'Fixes' tag
---
 kernel/bpf/core.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)

diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 13e9dbeeedf3..81f814f31187 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -1938,6 +1938,11 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
 		CONT;							\
 	LDX_MEM_##SIZEOP:						\
 		DST = *(SIZE *)(unsigned long) (SRC + insn->off);	\
+		CONT;							\
+	LDX_PROBE_MEM_##SIZEOP:						\
+		bpf_probe_read_kernel(&DST, sizeof(SIZE),		\
+				      (const void *)(long) (SRC + insn->off));	\
+		DST = *((SIZE *)&DST);					\
 		CONT;
 
 	LDST(B,   u8)
@@ -1945,15 +1950,6 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
 	LDST(W,  u32)
 	LDST(DW, u64)
 #undef LDST
-#define LDX_PROBE(SIZEOP, SIZE)							\
-	LDX_PROBE_MEM_##SIZEOP:							\
-		bpf_probe_read_kernel(&DST, SIZE, (const void *)(long) (SRC + insn->off));	\
-		CONT;
-	LDX_PROBE(B,  1)
-	LDX_PROBE(H,  2)
-	LDX_PROBE(W,  4)
-	LDX_PROBE(DW, 8)
-#undef LDX_PROBE
 
 #define ATOMIC_ALU_OP(BOP, KOP)						\
 		case BOP:						\
-- 
2.36.1

Hello:

This patch was applied to bpf/bpf.git (master)
by Daniel Borkmann <daniel@iogearbox.net>:

On Tue, 24 May 2022 10:12:27 +0800 you wrote:
> From: Menglong Dong <imagedong@tencent.com>
> 
> I think there is something wrong with BPF_PROBE_MEM in ___bpf_prog_run()
> in big-endian machine. Let's make a test and see what will happen if we
> want to load a 'u16' with BPF_PROBE_MEM.
> 
> Let's make the src value '0x0001', the value of dest register will become
> 0x0001000000000000, as the value will be loaded to the first 2 byte of
> DST with following code:
> 
> [...]

Here is the summary with links:
  - [bpf,v2] bpf: fix probe read error in ___bpf_prog_run()
    https://git.kernel.org/bpf/bpf/c/caff1fa4118c

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html