bpf, libbpf: Guard bpf inline asm from bpf_tail_call_static
authorDaniel Borkmann <daniel@iogearbox.net>
Wed, 21 Oct 2020 20:32:57 +0000 (22:32 +0200)
committerDaniel Borkmann <daniel@iogearbox.net>
Wed, 21 Oct 2020 23:46:52 +0000 (01:46 +0200)
Yaniv reported a compilation error after pulling latest libbpf:

  [...]
  ../libbpf/src/root/usr/include/bpf/bpf_helpers.h:99:10: error:
  unknown register name 'r0' in asm
                     : "r0", "r1", "r2", "r3", "r4", "r5");
  [...]

The issue got triggered given Yaniv was compiling tracing programs with native
target (e.g. x86) instead of BPF target, hence no BTF generated vmlinux.h nor
CO-RE used, and later llc with -march=bpf was invoked to compile from LLVM IR
to BPF object file. Given that clang was expecting x86 inline asm and not BPF
one the error complained that these regs don't exist on the former.

Guard bpf_tail_call_static() with defined(__bpf__) where BPF inline asm is valid
to use. BPF tracing programs on more modern kernels use BPF target anyway and
thus the bpf_tail_call_static() function will be available for them. BPF inline
asm is supported since clang 7 (clang <= 6 otherwise throws same above error),
and __bpf_unreachable() since clang 8, therefore include the latter condition
in order to prevent compilation errors for older clang versions. Given even an
old Ubuntu 18.04 LTS has official LLVM packages all the way up to llvm-10, I did
not bother to special case the __bpf_unreachable() inside bpf_tail_call_static()
further.

Also, undo the sockex3_kern's use of bpf_tail_call_static() sample given they
still have the old hacky way to even compile networking progs with native instead
of BPF target so bpf_tail_call_static() won't be defined there anymore.

Fixes: 0e9f6841f664 ("bpf, libbpf: Add bpf_tail_call_static helper for bpf programs")
Reported-by: Yaniv Agman <yanivagman@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Tested-by: Yaniv Agman <yanivagman@gmail.com>
Link: https://lore.kernel.org/bpf/CAMy7=ZUk08w5Gc2Z-EKi4JFtuUCaZYmE4yzhJjrExXpYKR4L8w@mail.gmail.com
Link: https://lore.kernel.org/bpf/20201021203257.26223-1-daniel@iogearbox.net
samples/bpf/sockex3_kern.c
tools/lib/bpf/bpf_helpers.h

index 8142d02b33e62f6113d6dddd23ec743c70d35fc9..b363503357e5dd8e861ed35f7070151e6033368f 100644 (file)
@@ -44,17 +44,17 @@ static inline void parse_eth_proto(struct __sk_buff *skb, u32 proto)
        switch (proto) {
        case ETH_P_8021Q:
        case ETH_P_8021AD:
-               bpf_tail_call_static(skb, &jmp_table, PARSE_VLAN);
+               bpf_tail_call(skb, &jmp_table, PARSE_VLAN);
                break;
        case ETH_P_MPLS_UC:
        case ETH_P_MPLS_MC:
-               bpf_tail_call_static(skb, &jmp_table, PARSE_MPLS);
+               bpf_tail_call(skb, &jmp_table, PARSE_MPLS);
                break;
        case ETH_P_IP:
-               bpf_tail_call_static(skb, &jmp_table, PARSE_IP);
+               bpf_tail_call(skb, &jmp_table, PARSE_IP);
                break;
        case ETH_P_IPV6:
-               bpf_tail_call_static(skb, &jmp_table, PARSE_IPV6);
+               bpf_tail_call(skb, &jmp_table, PARSE_IPV6);
                break;
        }
 }
index 2bdb7d6dbad21018a050e2e84df5a65de0ccf7e2..72b251110c4d75a8125a2659d8499ba7c525d59f 100644 (file)
@@ -72,6 +72,7 @@
 /*
  * Helper function to perform a tail call with a constant/immediate map slot.
  */
+#if __clang_major__ >= 8 && defined(__bpf__)
 static __always_inline void
 bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 {
@@ -98,6 +99,7 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
                     :: [ctx]"r"(ctx), [map]"r"(map), [slot]"i"(slot)
                     : "r0", "r1", "r2", "r3", "r4", "r5");
 }
+#endif
 
 /*
  * Helper structure used by eBPF C program