Re: Ask for help about bpf map

Re: Ask for help about bpf map

[Andrii Nakryiko]

On Mon, Jul 12, 2021 at 11:35 PM luwei (O) <luwei32@huawei.com> wrote:
>
> Hi, List:
>
>        I am a beginner about bpf and working on XDP now. I meet a
> problem and feel difficult to figure it out.
>
>        In my following codes, I use two ways to define my_map: in SEC
> maps and SEC .maps respectively. When I load the xdp_kern.o file,
>
> It has different results. The way I load is: ip link set dev ens3 xdp
> obj xdp1_kern.o sec xdp1.
>
>        when I define my_map using SEC maps, it loads successfully but
> fails to load using SEC .maps, it reports:
>
> "
>
> [12] TYPEDEF __u32 type_id=13
> [13] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
> [14] FUNC_PROTO (anon) return=2 args=(10 ctx)
> [15] FUNC xdp_prog1 type_id=14
> [16] INT char size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
> [17] ARRAY (anon) type_id=16 index_type_id=4 nr_elems=4
> [18] VAR _license type_id=17 linkage=1
> [19] DATASEC .maps size=0 vlen=1 size == 0
>
>
> Prog section 'xdp1' rejected: Permission denied (13)!
>   - Type:         6
>   - Instructions: 9 (0 over limit)
>   - License:      GPL
>
> Verifier analysis:
>
> 0: (b7) r1 = 0
> 1: (63) *(u32 *)(r10 -4) = r1
> last_idx 1 first_idx 0
> regs=2 stack=0 before 0: (b7) r1 = 0
> 2: (bf) r2 = r10
> 3: (07) r2 += -4
> 4: (18) r1 = 0x0

this shouldn't be 0x0.

I suspect you have an old iproute2 which doesn't yet use libbpf to
load BPF programs, so .maps definition is not yet supported. cc'ing
netdev@vger, David and Toke

> 6: (85) call bpf_map_lookup_elem#1
> R1 type=inv expected=map_ptr
> processed 6 insns (limit 1000000) max_states_per_insn 0 total_states 0
> peak_states 0 mark_read 0
>
> Error fetching program/map!
>
> "
>
> I tried to search google, and only found the following page:
>
> https://stackoverflow.com/questions/67553794/what-is-variable-attribute-sec-means
>
> Does anyone meet the same problem or help to explain this or provide me
> some suggestions ? Thanks !
>
>
> === xdp1_kern.c ===
>
> #define KBUILD_MODNAME "foo"
> #include <uapi/linux/bpf.h>
> #include <linux/time.h>
> #include <linux/in.h>
> #include <linux/if_ether.h>
> #include <linux/if_packet.h>
> #include <linux/if_vlan.h>
> #include <linux/ip.h>
> #include <linux/ipv6.h>
> #include <bpf/bpf_helpers.h>
>
> struct {
>      __uint(type, BPF_MAP_TYPE_HASH);
>      __uint(max_entries, 1024);
>      __type(key, int);
>      __type(value, int);
> } my_map SEC(".maps");
>
> #if 0
> #define PIN_GLOBAL_NS           2
> struct bpf_elf_map {
>          __u32 type;
>          __u32 size_key;
>          __u32 size_value;
>          __u32 max_elem;
>          __u32 flags;
>          __u32 id;
>          __u32 pinning;
> };
>
> struct bpf_elf_map SEC("maps") my_map = {
>          .type = BPF_MAP_TYPE_HASH,
>          .size_key = sizeof(int),
>          .size_value = sizeof(int),
>          .pinning        = PIN_GLOBAL_NS,
>          .max_elem = 65535,
> };
> #endif
>
> SEC("xdp1")
> int xdp_prog1(struct xdp_md *ctx)
> {
>      int key = 0;
>      struct map_elem *val;
>
>      val = bpf_map_lookup_elem(&my_map, &key);
>      if (val) {
>          return XDP_PASS;
>      }
>
>      return XDP_PASS;
> }
>
> char _license[] SEC("license") = "GPL";
>
> --
> Best Regards,
> Lu Wei
>

Re: Ask for help about bpf map

[Toke Høiland-Jørgensen]

Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:

> On Mon, Jul 12, 2021 at 11:35 PM luwei (O) <luwei32@huawei.com> wrote:
>>
>> Hi, List:
>>
>>        I am a beginner about bpf and working on XDP now. I meet a
>> problem and feel difficult to figure it out.
>>
>>        In my following codes, I use two ways to define my_map: in SEC
>> maps and SEC .maps respectively. When I load the xdp_kern.o file,
>>
>> It has different results. The way I load is: ip link set dev ens3 xdp
>> obj xdp1_kern.o sec xdp1.
>>
>>        when I define my_map using SEC maps, it loads successfully but
>> fails to load using SEC .maps, it reports:
>>
>> "
>>
>> [12] TYPEDEF __u32 type_id=13
>> [13] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
>> [14] FUNC_PROTO (anon) return=2 args=(10 ctx)
>> [15] FUNC xdp_prog1 type_id=14
>> [16] INT char size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
>> [17] ARRAY (anon) type_id=16 index_type_id=4 nr_elems=4
>> [18] VAR _license type_id=17 linkage=1
>> [19] DATASEC .maps size=0 vlen=1 size == 0
>>
>>
>> Prog section 'xdp1' rejected: Permission denied (13)!
>>   - Type:         6
>>   - Instructions: 9 (0 over limit)
>>   - License:      GPL
>>
>> Verifier analysis:
>>
>> 0: (b7) r1 = 0
>> 1: (63) *(u32 *)(r10 -4) = r1
>> last_idx 1 first_idx 0
>> regs=2 stack=0 before 0: (b7) r1 = 0
>> 2: (bf) r2 = r10
>> 3: (07) r2 += -4
>> 4: (18) r1 = 0x0
>
> this shouldn't be 0x0.
>
> I suspect you have an old iproute2 which doesn't yet use libbpf to
> load BPF programs, so .maps definition is not yet supported. cc'ing
> netdev@vger, David and Toke

That would be my guess as well; what's the output of 'ip -V'?

-Toke

Re: Ask for help about bpf map

[luwei (O)]

I have updated the iproute2 according this page: 
https://github.com/cilium/cilium/issues/7446

Now I use this version of iproute2: https://github.com/shemminger/iproute2

The version of iproute2 is 5.11, and the kernel version is 5.13(the 
latest version).


在 2021/7/14 1:07 AM, Toke Høiland-Jørgensen 写道:
> Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:
>
>> On Mon, Jul 12, 2021 at 11:35 PM luwei (O) <luwei32@huawei.com> wrote:
>>> Hi, List:
>>>
>>>         I am a beginner about bpf and working on XDP now. I meet a
>>> problem and feel difficult to figure it out.
>>>
>>>         In my following codes, I use two ways to define my_map: in SEC
>>> maps and SEC .maps respectively. When I load the xdp_kern.o file,
>>>
>>> It has different results. The way I load is: ip link set dev ens3 xdp
>>> obj xdp1_kern.o sec xdp1.
>>>
>>>         when I define my_map using SEC maps, it loads successfully but
>>> fails to load using SEC .maps, it reports:
>>>
>>> "
>>>
>>> [12] TYPEDEF __u32 type_id=13
>>> [13] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
>>> [14] FUNC_PROTO (anon) return=2 args=(10 ctx)
>>> [15] FUNC xdp_prog1 type_id=14
>>> [16] INT char size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
>>> [17] ARRAY (anon) type_id=16 index_type_id=4 nr_elems=4
>>> [18] VAR _license type_id=17 linkage=1
>>> [19] DATASEC .maps size=0 vlen=1 size == 0
>>>
>>>
>>> Prog section 'xdp1' rejected: Permission denied (13)!
>>>    - Type:         6
>>>    - Instructions: 9 (0 over limit)
>>>    - License:      GPL
>>>
>>> Verifier analysis:
>>>
>>> 0: (b7) r1 = 0
>>> 1: (63) *(u32 *)(r10 -4) = r1
>>> last_idx 1 first_idx 0
>>> regs=2 stack=0 before 0: (b7) r1 = 0
>>> 2: (bf) r2 = r10
>>> 3: (07) r2 += -4
>>> 4: (18) r1 = 0x0
>> this shouldn't be 0x0.
>>
>> I suspect you have an old iproute2 which doesn't yet use libbpf to
>> load BPF programs, so .maps definition is not yet supported. cc'ing
>> netdev@vger, David and Toke
> That would be my guess as well; what's the output of 'ip -V'?
>
> -Toke
>
> .

-- 
Best Regards,
Lu Wei

Re: Ask for help about bpf map

[luwei (O)]

I tried 5.13 version in this page: 
https://git.kernel.org/pub/scm/network/iproute2/iproute2.git , still 
failed with the same error.

在 2021/7/14 9:05 AM, luwei (O) 写道:
> I have updated the iproute2 according this page: 
> https://github.com/cilium/cilium/issues/7446
>
> Now I use this version of iproute2: 
> https://github.com/shemminger/iproute2
>
> The version of iproute2 is 5.11, and the kernel version is 5.13(the 
> latest version).
>
>
> 在 2021/7/14 1:07 AM, Toke Høiland-Jørgensen 写道:
>> Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:
>>
>>> On Mon, Jul 12, 2021 at 11:35 PM luwei (O) <luwei32@huawei.com> wrote:
>>>> Hi, List:
>>>>
>>>>         I am a beginner about bpf and working on XDP now. I meet a
>>>> problem and feel difficult to figure it out.
>>>>
>>>>         In my following codes, I use two ways to define my_map: in SEC
>>>> maps and SEC .maps respectively. When I load the xdp_kern.o file,
>>>>
>>>> It has different results. The way I load is: ip link set dev ens3 xdp
>>>> obj xdp1_kern.o sec xdp1.
>>>>
>>>>         when I define my_map using SEC maps, it loads successfully but
>>>> fails to load using SEC .maps, it reports:
>>>>
>>>> "
>>>>
>>>> [12] TYPEDEF __u32 type_id=13
>>>> [13] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
>>>> [14] FUNC_PROTO (anon) return=2 args=(10 ctx)
>>>> [15] FUNC xdp_prog1 type_id=14
>>>> [16] INT char size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
>>>> [17] ARRAY (anon) type_id=16 index_type_id=4 nr_elems=4
>>>> [18] VAR _license type_id=17 linkage=1
>>>> [19] DATASEC .maps size=0 vlen=1 size == 0
>>>>
>>>>
>>>> Prog section 'xdp1' rejected: Permission denied (13)!
>>>>    - Type:         6
>>>>    - Instructions: 9 (0 over limit)
>>>>    - License:      GPL
>>>>
>>>> Verifier analysis:
>>>>
>>>> 0: (b7) r1 = 0
>>>> 1: (63) *(u32 *)(r10 -4) = r1
>>>> last_idx 1 first_idx 0
>>>> regs=2 stack=0 before 0: (b7) r1 = 0
>>>> 2: (bf) r2 = r10
>>>> 3: (07) r2 += -4
>>>> 4: (18) r1 = 0x0
>>> this shouldn't be 0x0.
>>>
>>> I suspect you have an old iproute2 which doesn't yet use libbpf to
>>> load BPF programs, so .maps definition is not yet supported. cc'ing
>>> netdev@vger, David and Toke
>> That would be my guess as well; what's the output of 'ip -V'?
>>
>> -Toke
>>
>> .
>
-- 
Best Regards,
Lu Wei

Re: Ask for help about bpf map

[luwei (O)]

Hi Andrii and toke,

     I have sovled this issue. The reason is that my iproute2 does not 
support libbpf, once I compile iproute2 with libbpf, it works. Thanks 
for reply!

在 2021/7/14 10:02 AM, luwei (O) 写道:
> I tried 5.13 version in this page: 
> https://git.kernel.org/pub/scm/network/iproute2/iproute2.git , still 
> failed with the same error.
>
> 在 2021/7/14 9:05 AM, luwei (O) 写道:
>> I have updated the iproute2 according this page: 
>> https://github.com/cilium/cilium/issues/7446
>>
>> Now I use this version of iproute2: 
>> https://github.com/shemminger/iproute2
>>
>> The version of iproute2 is 5.11, and the kernel version is 5.13(the 
>> latest version).
>>
>>
>> 在 2021/7/14 1:07 AM, Toke Høiland-Jørgensen 写道:
>>> Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:
>>>
>>>> On Mon, Jul 12, 2021 at 11:35 PM luwei (O) <luwei32@huawei.com> wrote:
>>>>> Hi, List:
>>>>>
>>>>>         I am a beginner about bpf and working on XDP now. I meet a
>>>>> problem and feel difficult to figure it out.
>>>>>
>>>>>         In my following codes, I use two ways to define my_map: in 
>>>>> SEC
>>>>> maps and SEC .maps respectively. When I load the xdp_kern.o file,
>>>>>
>>>>> It has different results. The way I load is: ip link set dev ens3 xdp
>>>>> obj xdp1_kern.o sec xdp1.
>>>>>
>>>>>         when I define my_map using SEC maps, it loads successfully 
>>>>> but
>>>>> fails to load using SEC .maps, it reports:
>>>>>
>>>>> "
>>>>>
>>>>> [12] TYPEDEF __u32 type_id=13
>>>>> [13] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
>>>>> [14] FUNC_PROTO (anon) return=2 args=(10 ctx)
>>>>> [15] FUNC xdp_prog1 type_id=14
>>>>> [16] INT char size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
>>>>> [17] ARRAY (anon) type_id=16 index_type_id=4 nr_elems=4
>>>>> [18] VAR _license type_id=17 linkage=1
>>>>> [19] DATASEC .maps size=0 vlen=1 size == 0
>>>>>
>>>>>
>>>>> Prog section 'xdp1' rejected: Permission denied (13)!
>>>>>    - Type:         6
>>>>>    - Instructions: 9 (0 over limit)
>>>>>    - License:      GPL
>>>>>
>>>>> Verifier analysis:
>>>>>
>>>>> 0: (b7) r1 = 0
>>>>> 1: (63) *(u32 *)(r10 -4) = r1
>>>>> last_idx 1 first_idx 0
>>>>> regs=2 stack=0 before 0: (b7) r1 = 0
>>>>> 2: (bf) r2 = r10
>>>>> 3: (07) r2 += -4
>>>>> 4: (18) r1 = 0x0
>>>> this shouldn't be 0x0.
>>>>
>>>> I suspect you have an old iproute2 which doesn't yet use libbpf to
>>>> load BPF programs, so .maps definition is not yet supported. cc'ing
>>>> netdev@vger, David and Toke
>>> That would be my guess as well; what's the output of 'ip -V'?
>>>
>>> -Toke
>>>
>>> .
>>
-- 
Best Regards,
Lu Wei

Re: Ask for help about bpf map

[Toke Høiland-Jørgensen]

"luwei (O)" <luwei32@huawei.com> writes:

> Hi Andrii and toke,
>
>      I have sovled this issue. The reason is that my iproute2 does not 
> support libbpf, once I compile iproute2 with libbpf, it works. Thanks 
> for reply!

Awesome! You're welcome :)

-Toke

Re: Ask for help about bpf map

[Alexei Starovoitov]

On Wed, Jul 14, 2021 at 1:24 AM luwei (O) <luwei32@huawei.com> wrote:
>
> Hi Andrii and toke,
>
>      I have sovled this issue. The reason is that my iproute2 does not
> support libbpf, once I compile iproute2 with libbpf, it works. Thanks
> for reply!

How did you figure that out?
I thought iproute folks should have included that info as part of -V output.
Since this exact concern was hotly debated in the past.
Non-vendoring clearly causes this annoying user experience.

Re: Ask for help about bpf map

[luwei (O)]

Andrii and Toke inspired me. You are right, the libbpf version should be included in -V output
, but not mine. I searched google and found this page: https://www.spinics.net/lists/netdev/msg700482.html
, according which I re-compiled iproute2 and it works.

Previously, I downloaded iproute2(5.13) and excuted "cd iproute2; make && make install". Libbpf which supports
btf-defined map is not included.


在 2021/7/14 10:48 PM, Alexei Starovoitov 写道:
> On Wed, Jul 14, 2021 at 1:24 AM luwei (O) <luwei32@huawei.com> wrote:
>> Hi Andrii and toke,
>>
>>       I have sovled this issue. The reason is that my iproute2 does not
>> support libbpf, once I compile iproute2 with libbpf, it works. Thanks
>> for reply!
> How did you figure that out?
> I thought iproute folks should have included that info as part of -V output.
> Since this exact concern was hotly debated in the past.
> Non-vendoring clearly causes this annoying user experience.
> .

-- 
Best Regards,
Lu Wei

Re: Ask for help about bpf map

[Toke Høiland-Jørgensen]

"luwei (O)" <luwei32@huawei.com> writes:

> Andrii and Toke inspired me. You are right, the libbpf version should be included in -V output
> , but not mine. I searched google and found this page: https://www.spinics.net/lists/netdev/msg700482.html
> , according which I re-compiled iproute2 and it works.

Did the libbpf version appear in the output of 'ip -V' after you
recompiled and enabled it? It does in mine:

$ ./ip/ip -V
ip utility, iproute2-5.13.0, libbpf 0.4.0

-Toke

Re: Ask for help about bpf map

[luwei (O)]

It's very strange, in my virtual host, it is:

$ ip -V

ip utility, iproute2-5.11.0


but in my physical host:

$ ip -V
ip utility, iproute2-5.11.0, libbpf 0.5.0


I compiled iproute2 in the same way as I mentioned previously, and the 
kernel versions are both 5.13 (in fact the same code) .


在 2021/7/19 8:38 PM, Toke Høiland-Jørgensen 写道:
> "luwei (O)" <luwei32@huawei.com> writes:
>
>> Andrii and Toke inspired me. You are right, the libbpf version should be included in -V output
>> , but not mine. I searched google and found this page: https://www.spinics.net/lists/netdev/msg700482.html
>> , according which I re-compiled iproute2 and it works.
> Did the libbpf version appear in the output of 'ip -V' after you
> recompiled and enabled it? It does in mine:
>
> $ ./ip/ip -V
> ip utility, iproute2-5.13.0, libbpf 0.4.0
>
> -Toke
>
> .

-- 
Best Regards,
Lu Wei

Re: Ask for help about bpf map

[Toke Høiland-Jørgensen]

"luwei (O)" <luwei32@huawei.com> writes:

> It's very strange, in my virtual host, it is:
>
> $ ip -V
>
> ip utility, iproute2-5.11.0
>
>
> but in my physical host:
>
> $ ip -V
> ip utility, iproute2-5.11.0, libbpf 0.5.0
>
>
> I compiled iproute2 in the same way as I mentioned previously, and the 
> kernel versions are both 5.13 (in fact the same code) .

When compiling, the configure script should tell you whether it can find
libbpf. If not, it's probably because you don't have the right header
files installed...

-Toke

Re: Ask for help about bpf map

[luwei (O)]

OK, I will check that later.

在 2021/7/20 10:50 PM, Toke Høiland-Jørgensen 写道:
> "luwei (O)" <luwei32@huawei.com> writes:
>
>> It's very strange, in my virtual host, it is:
>>
>> $ ip -V
>>
>> ip utility, iproute2-5.11.0
>>
>>
>> but in my physical host:
>>
>> $ ip -V
>> ip utility, iproute2-5.11.0, libbpf 0.5.0
>>
>>
>> I compiled iproute2 in the same way as I mentioned previously, and the
>> kernel versions are both 5.13 (in fact the same code) .
> When compiling, the configure script should tell you whether it can find
> libbpf. If not, it's probably because you don't have the right header
> files installed...
>
> -Toke
>
> .

-- 
Best Regards,
Lu Wei