Re: [PATCH ipsec-next v1 6/7] bpf: selftests: test_tunnel: Disable CO-RE relocations

[Yonghong Song <yonghong.song@linux.dev>]

On 11/27/23 7:01 PM, Daniel Xu wrote:
> On Mon, Nov 27, 2023 at 02:45:11PM -0600, Daniel Xu wrote:
>> On Sun, Nov 26, 2023 at 09:53:04PM -0800, Yonghong Song wrote:
>>> On 11/27/23 12:44 AM, Yonghong Song wrote:
>>>> On 11/26/23 8:52 PM, Eduard Zingerman wrote:
>>>>> On Sun, 2023-11-26 at 18:04 -0600, Daniel Xu wrote:
>>>>> [...]
>>>>>>> Tbh I'm not sure. This test passes with preserve_static_offset
>>>>>>> because it suppresses preserve_access_index. In general clang
>>>>>>> translates bitfield access to a set of IR statements like:
>>>>>>>
>>>>>>>     C:
>>>>>>>       struct foo {
>>>>>>>         unsigned _;
>>>>>>>         unsigned a:1;
>>>>>>>         ...
>>>>>>>       };
>>>>>>>       ... foo->a ...
>>>>>>>
>>>>>>>     IR:
>>>>>>>       %a = getelementptr inbounds %struct.foo, ptr %0, i32 0, i32 1
>>>>>>>       %bf.load = load i8, ptr %a, align 4
>>>>>>>       %bf.clear = and i8 %bf.load, 1
>>>>>>>       %bf.cast = zext i8 %bf.clear to i32
>>>>>>>
>>>>>>> With preserve_static_offset the getelementptr+load are replaced by a
>>>>>>> single statement which is preserved as-is till code generation,
>>>>>>> thus load with align 4 is preserved.
>>>>>>>
>>>>>>> On the other hand, I'm not sure that clang guarantees that load or
>>>>>>> stores used for bitfield access would be always aligned according to
>>>>>>> verifier expectations.
>>>>>>>
>>>>>>> I think we should check if there are some clang knobs that prevent
>>>>>>> generation of unaligned memory access. I'll take a look.
>>>>>> Is there a reason to prefer fixing in compiler? I'm not opposed to it,
>>>>>> but the downside to compiler fix is it takes years to propagate and
>>>>>> sprinkles ifdefs into the code.
>>>>>>
>>>>>> Would it be possible to have an analogue of BPF_CORE_READ_BITFIELD()?
>>>>> Well, the contraption below passes verification, tunnel selftest
>>>>> appears to work. I might have messed up some shifts in the macro,
>>>>> though.
>>>> I didn't test it. But from high level it should work.
>>>>
>>>>> Still, if clang would peek unlucky BYTE_{OFFSET,SIZE} for a particular
>>>>> field access might be unaligned.
>>>> clang should pick a sensible BYTE_SIZE/BYTE_OFFSET to meet
>>>> alignment requirement. This is also required for BPF_CORE_READ_BITFIELD.
>>>>
>>>>> ---
>>>>>
>>>>> diff --git a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
>>>>> b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
>>>>> index 3065a716544d..41cd913ac7ff 100644
>>>>> --- a/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
>>>>> +++ b/tools/testing/selftests/bpf/progs/test_tunnel_kern.c
>>>>> @@ -9,6 +9,7 @@
>>>>>    #include "vmlinux.h"
>>>>>    #include <bpf/bpf_helpers.h>
>>>>>    #include <bpf/bpf_endian.h>
>>>>> +#include <bpf/bpf_core_read.h>
>>>>>    #include "bpf_kfuncs.h"
>>>>>    #include "bpf_tracing_net.h"
>>>>>    @@ -144,6 +145,38 @@ int ip6gretap_get_tunnel(struct __sk_buff *skb)
>>>>>        return TC_ACT_OK;
>>>>>    }
>>>>>    +#define BPF_CORE_WRITE_BITFIELD(s, field, new_val) ({            \
>>>>> +    void *p = (void *)s + __CORE_RELO(s, field, BYTE_OFFSET);    \
>>>>> +    unsigned byte_size = __CORE_RELO(s, field, BYTE_SIZE);        \
>>>>> +    unsigned lshift = __CORE_RELO(s, field, LSHIFT_U64); \
>>>>> +    unsigned rshift = __CORE_RELO(s, field, RSHIFT_U64); \
>>>>> +    unsigned bit_size = (rshift - lshift);                \
>>>>> +    unsigned long long nval, val, hi, lo;                \
>>>>> +                                    \
>>>>> +    asm volatile("" : "=r"(p) : "0"(p));                \
>>>> Use asm volatile("" : "+r"(p)) ?
>>>>
>>>>> +                                    \
>>>>> +    switch (byte_size) {                        \
>>>>> +    case 1: val = *(unsigned char *)p; break;            \
>>>>> +    case 2: val = *(unsigned short *)p; break;            \
>>>>> +    case 4: val = *(unsigned int *)p; break;            \
>>>>> +    case 8: val = *(unsigned long long *)p; break;            \
>>>>> +    }                                \
>>>>> +    hi = val >> (bit_size + rshift);                \
>>>>> +    hi <<= bit_size + rshift;                    \
>>>>> +    lo = val << (bit_size + lshift);                \
>>>>> +    lo >>= bit_size + lshift;                    \
>>>>> +    nval = new_val;                            \
>>>>> +    nval <<= lshift;                        \
>>>>> +    nval >>= rshift;                        \
>>>>> +    val = hi | nval | lo;                        \
>>>>> +    switch (byte_size) {                        \
>>>>> +    case 1: *(unsigned char *)p      = val; break;            \
>>>>> +    case 2: *(unsigned short *)p     = val; break;            \
>>>>> +    case 4: *(unsigned int *)p       = val; break;            \
>>>>> +    case 8: *(unsigned long long *)p = val; break;            \
>>>>> +    }                                \
>>>>> +})
>>>> I think this should be put in libbpf public header files but not sure
>>>> where to put it. bpf_core_read.h although it is core write?
>>>>
>>>> But on the other hand, this is a uapi struct bitfield write,
>>>> strictly speaking, CORE write is really unnecessary here. It
>>>> would be great if we can relieve users from dealing with
>>>> such unnecessary CORE writes. In that sense, for this particular
>>>> case, I would prefer rewriting the code by using byte-level
>>>> stores...
>>> or preserve_static_offset to clearly mean to undo bitfield CORE ...
>> Ok, I will do byte-level rewrite for next revision.
> [...]
>
> This patch seems to work: https://pastes.dxuuu.xyz/0glrf9 .
>
> But I don't think it's very pretty. Also I'm seeing on the internet that
> people are saying the exact layout of bitfields is compiler dependent.

Any reference for this (exact layout of bitfields is compiler dependent)?

> So I am wondering if these byte sized writes are correct. For that
> matter, I am wondering how the GCC generated bitfield accesses line up
> with clang generated BPF bytecode. Or why uapi contains a bitfield.

One thing for sure is memory layout of bitfields should be the same
for both clang and gcc as it is determined by C standard. Register
representation and how to manipulate could be different for different
compilers.

>
> WDYT, should I send up v2 with this or should I do one of the other
> approaches in this thread?

Daniel, look at your patch, since we need to do CORE_READ for
those bitfields any way, I think Eduard's patch with
BPF_CORE_WRITE_BITFIELD does make sense and it also makes code
easy to understand. Could you take Eduard's patch for now?
Whether and where to put BPF_CORE_WRITE_BITFIELD macros
can be decided later.

>
> I am ok with any of the approaches.
>
> Thanks,
> Daniel
>