Re: [bpf/tools] cd17d77705: kernel_selftests.bpf.test_sock_addr.sh.fail

From: Andrii Nakryiko
Date: Thu Jun 27 2019 - 21:56:55 EST

Next message: Thiago Jung Bauermann: "Re: [RFC PATCH] virtio_ring: Use DMA API if guest memory is encrypted"
Previous message: Shakeel Butt: "[PATCH] mm, vmscan: prevent useless kswapd loops"
In reply to: Stanislav Fomichev: "Re: [bpf/tools] cd17d77705: kernel_selftests.bpf.test_sock_addr.sh.fail"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Thu, Jun 27, 2019 at 8:50 AM Stanislav Fomichev <sdf@xxxxxxxxxxx> wrote:
>
> On 06/27, kernel test robot wrote:
> > FYI, we noticed the following commit (built with gcc-7):
> >
> > commit: cd17d77705780e2270937fb3cbd2b985adab3edc ("bpf/tools: sync bpf.h")
> > https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master
> >
> > in testcase: kernel_selftests
> > with following parameters:
> >
> > group: kselftests-00
> >
> > test-description: The kernel contains a set of "self tests" under the tools/testing/selftests/ directory. These are intended to be small unit tests to exercise individual code paths in the kernel.
> > test-url: https://www.kernel.org/doc/Documentation/kselftest.txt
> >
> >
> > on test machine: qemu-system-x86_64 -enable-kvm -cpu SandyBridge -smp 2 -m 8G
> >
> > caused below changes (please refer to attached dmesg/kmsg for entire log/backtrace):
> >
> > # ; int connect_v6_prog(struct bpf_sock_addr *ctx)
> > # 0: (bf) r6 = r1
> > # 1: (18) r1 = 0x100000000000000
> > # ; tuple.ipv6.daddr[0] = bpf_htonl(DST_REWRITE_IP6_0);
> > # 3: (7b) *(u64 *)(r10 -16) = r1
> > # 4: (b7) r1 = 169476096
> > # ; memset(&tuple.ipv6.sport, 0, sizeof(tuple.ipv6.sport));
> > # 5: (63) *(u32 *)(r10 -8) = r1
> > # 6: (b7) r7 = 0
> > # ; tuple.ipv6.daddr[0] = bpf_htonl(DST_REWRITE_IP6_0);
> > # 7: (7b) *(u64 *)(r10 -24) = r7
> > # 8: (7b) *(u64 *)(r10 -32) = r7
> > # 9: (7b) *(u64 *)(r10 -40) = r7
> > # ; if (ctx->type != SOCK_STREAM && ctx->type != SOCK_DGRAM)
> > # 10: (61) r1 = *(u32 *)(r6 +32)
> > # ; if (ctx->type != SOCK_STREAM && ctx->type != SOCK_DGRAM)
> > # 11: (bf) r2 = r1
> > # 12: (07) r2 += -1
> > # 13: (67) r2 <<= 32
> > # 14: (77) r2 >>= 32
> > # 15: (25) if r2 > 0x1 goto pc+33
> > # R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=1,var_off=(0x0; 0x1)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=00000000 fp-32=00000000 fp-40=00000000
> > # ; else if (ctx->type == SOCK_STREAM)
> > # 16: (55) if r1 != 0x1 goto pc+8
> > # R1=inv1 R2=inv(id=0,umax_value=1,var_off=(0x0; 0x1)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=00000000 fp-32=00000000 fp-40=00000000
> > # 17: (bf) r2 = r10
> > # ; sk = bpf_sk_lookup_tcp(ctx, &tuple, sizeof(tuple.ipv6),
> > # 18: (07) r2 += -40
> > # 19: (bf) r1 = r6
> > # 20: (b7) r3 = 36
> > # 21: (b7) r4 = -1
> > # 22: (b7) r5 = 0
> > # 23: (85) call bpf_sk_lookup_tcp#84
> > # 24: (05) goto pc+7
> > # ; if (!sk)
> > # 32: (15) if r0 == 0x0 goto pc+16
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; if (sk->src_ip6[0] != tuple.ipv6.daddr[0] ||
> > # 33: (61) r1 = *(u32 *)(r0 +28)
> > # ; if (sk->src_ip6[0] != tuple.ipv6.daddr[0] ||
> > # 34: (61) r2 = *(u32 *)(r10 -24)
> > # ; if (sk->src_ip6[0] != tuple.ipv6.daddr[0] ||
> > # 35: (5d) if r1 != r2 goto pc+11
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; sk->src_ip6[1] != tuple.ipv6.daddr[1] ||
> > # 36: (61) r1 = *(u32 *)(r0 +32)
> > # ; sk->src_ip6[1] != tuple.ipv6.daddr[1] ||
> > # 37: (61) r2 = *(u32 *)(r10 -20)
> > # ; sk->src_ip6[1] != tuple.ipv6.daddr[1] ||
> > # 38: (5d) if r1 != r2 goto pc+8
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; sk->src_ip6[2] != tuple.ipv6.daddr[2] ||
> > # 39: (61) r1 = *(u32 *)(r0 +36)
> > # ; sk->src_ip6[2] != tuple.ipv6.daddr[2] ||
> > # 40: (61) r2 = *(u32 *)(r10 -16)
> > # ; sk->src_ip6[2] != tuple.ipv6.daddr[2] ||
> > # 41: (5d) if r1 != r2 goto pc+5
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; sk->src_ip6[3] != tuple.ipv6.daddr[3] ||
> > # 42: (61) r1 = *(u32 *)(r0 +40)
> > # ; sk->src_ip6[3] != tuple.ipv6.daddr[3] ||
> > # 43: (61) r2 = *(u32 *)(r10 -12)
> > # ; sk->src_ip6[3] != tuple.ipv6.daddr[3] ||
> > # 44: (5d) if r1 != r2 goto pc+2
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; sk->src_port != DST_REWRITE_PORT6) {
> > # 45: (61) r1 = *(u32 *)(r0 +44)
> > # ; if (sk->src_ip6[0] != tuple.ipv6.daddr[0] ||
> > # 46: (15) if r1 == 0x1a0a goto pc+4
> > # R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; bpf_sk_release(sk);
> > # 47: (bf) r1 = r0
> > # 48: (85) call bpf_sk_release#86
> > # ; }
> > # 49: (bf) r0 = r7
> > # 50: (95) exit
> > #
> > # from 46 to 51: R0=sock(id=0,ref_obj_id=2,off=0,imm=0) R1=inv6666 R2=inv(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R6=ctx(id=0,off=0,imm=0) R7=inv0 R10=fp0,call_-1 fp-8=????mmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm refs=2
> > # ; bpf_sk_release(sk);
> > # 51: (bf) r1 = r0
> > # 52: (85) call bpf_sk_release#86
> > # 53: (b7) r1 = 2586
> > # ; ctx->user_port = bpf_htons(DST_REWRITE_PORT6);
> > # 54: (63) *(u32 *)(r6 +24) = r1
> > # 55: (18) r1 = 0x100000000000000
> > # ; ctx->user_ip6[2] = bpf_htonl(DST_REWRITE_IP6_2);
> > # 57: (7b) *(u64 *)(r6 +16) = r1
> > # invalid bpf_context access off=16 size=8
> This looks like clang doing single u64 write for user_ip6[2] and
> user_ip6[3] instead of two u32. I don't think we allow that.
>
> I've seen this a couple of times myself while playing with some
> progs, but not sure what's the right way to 'fix' it.

So I bisected this to the commit below. That change effectively makes
`struct bpf_sock_addr` 8-byte aligned, which apparently triggers some
optimization in clang to generate this 8-byte write for two
consecutive 4-byte array elements.

commit cd17d77705780e2270937fb3cbd2b985adab3edc (HEAD, refs/bisect/bad)
Author: Stanislav Fomichev <sdf@xxxxxxxxxx>
Date: Wed Jun 12 10:30:39 2019 -0700

bpf/tools: sync bpf.h

Add sk to struct bpf_sock_addr and struct bpf_sock_ops.

Cc: Martin Lau <kafai@xxxxxx>
Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
Signed-off-by: Daniel Borkmann <daniel@xxxxxxxxxxxxx>

diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index ae0907d8c03a..d0a23476f887 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -3247,6 +3247,7 @@ struct bpf_sock_addr {
__u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
* Stored in network byte order.
*/
+ __bpf_md_ptr(struct bpf_sock *, sk);
};

/* User bpf_sock_ops struct to access socket values and specify request ops
@@ -3298,6 +3299,7 @@ struct bpf_sock_ops {
__u32 sk_txhash;
__u64 bytes_received;
__u64 bytes_acked;
+ __bpf_md_ptr(struct bpf_sock *, sk);
};

/* Definitions for bpf_sock_ops_cb_flags */

>
> > # processed 49 insns (limit 1000000) max_states_per_insn 0 total_states 13 peak_states 13 mark_read 11
> > #
> > # libbpf: -- END LOG --
> > # libbpf: failed to load program 'cgroup/connect6'
> > # libbpf: failed to load object './connect6_prog.o'
> > # (test_sock_addr.c:752: errno: Bad file descriptor) >>> Loading program (./connect6_prog.o) error.

Next message: Thiago Jung Bauermann: "Re: [RFC PATCH] virtio_ring: Use DMA API if guest memory is encrypted"
Previous message: Shakeel Butt: "[PATCH] mm, vmscan: prevent useless kswapd loops"
In reply to: Stanislav Fomichev: "Re: [bpf/tools] cd17d77705: kernel_selftests.bpf.test_sock_addr.sh.fail"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]