Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1663401 > unrolled thread

Re: [PATCH v2] arm: eBPF JIT compiler

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH v2] arm: eBPF JIT compiler Daniel Borkmann <daniel@iogearbox.net> - 2017-06-12 12:30 +0200
    Re: [PATCH v2] arm: eBPF JIT compiler Russell King - ARM Linux <linux@armlinux.org.uk> - 2017-06-12 13:10 +0200
      Re: [PATCH v2] arm: eBPF JIT compiler Shubham Bansal <illusionist.neo@gmail.com> - 2017-06-12 17:50 +0200
    Re: [PATCH v2] arm: eBPF JIT compiler Shubham Bansal <illusionist.neo@gmail.com> - 2017-06-12 17:50 +0200
      Re: [PATCH v2] arm: eBPF JIT compiler Alexander Alemayhu <alexander@alemayhu.com> - 2017-06-13 00:50 +0200
        Re: [PATCH v2] arm: eBPF JIT compiler David Miller <davem@davemloft.net> - 2017-06-13 00:50 +0200
      Re: [PATCH v2] arm: eBPF JIT compiler Daniel Borkmann <daniel@iogearbox.net> - 2017-06-13 01:20 +0200
      Re: [PATCH v2] arm: eBPF JIT compiler Shubham Bansal <illusionist.neo@gmail.com> - 2017-06-13 09:00 +0200
        Re: [PATCH v2] arm: eBPF JIT compiler Daniel Borkmann <daniel@iogearbox.net> - 2017-06-14 22:40 +0200
          Re: [PATCH v2] arm: eBPF JIT compiler Shubham Bansal <illusionist.neo@gmail.com> - 2017-06-17 14:30 +0200
            Re: [PATCH v2] arm: eBPF JIT compiler Daniel Borkmann <daniel@iogearbox.net> - 2017-06-19 20:20 +0200

#1663401 — Re: [PATCH v2] arm: eBPF JIT compiler

On 05/30/2017 09:19 PM, Kees Cook wrote:
> Forwarding this to net-dev and eBPF folks, who weren't on CC...

Sorry for being late. Some comments below from a cursory look ...

> -Kees
>
> On Thu, May 25, 2017 at 4:13 PM, Shubham Bansal
> <illusionist.neo@gmail.com> wrote:
>> The JIT compiler emits ARM 32 bit instructions. Currently, It supports
>> eBPF only. Classic BPF is supported because of the conversion by BPF
>> core.
>>
>> This patch is essentially changing the current implementation of JIT
>> compiler of Berkeley Packet Filter from classic to internal with almost
>> all instructions from eBPF ISA supported except the following
>>          BPF_ALU64 | BPF_DIV | BPF_K
>>          BPF_ALU64 | BPF_DIV | BPF_X
>>          BPF_ALU64 | BPF_MOD | BPF_K
>>          BPF_ALU64 | BPF_MOD | BPF_X
>>          BPF_STX | BPF_XADD | BPF_W
>>          BPF_STX | BPF_XADD | BPF_DW
>>          BPF_JMP | BPF_CALL

Any plans to implement above especially BPF_JMP | BPF_CALL in near future?
Reason why I'm asking is that i) currently the arm32 cBPF JIT implements
all of the cBPF extensions (except SKF_AD_RANDOM and SKF_AD_VLAN_TPID).
Some of the programs that were JITed before e.g. using SKF_AD_CPU would now
fall back to the eBPF interpreter due to lack of translation in JIT, but
also ii) that probably most (if not all) of eBPF programs use BPF helper
calls heavily, which will still redirect them to the interpreter right now
due to lack of BPF_JMP | BPF_CALL support, so it's really quite essential
to have it implemented.

>> Implementation is using scratch space to emulate 64 bit eBPF ISA on 32 bit
>> ARM because of deficiency of general purpose registers on ARM. Currently,
>> only LITTLE ENDIAN machines are supported in this eBPF JIT Compiler.
>>
>> Tested on ARMv7 with QEMU by me (Shubham Bansal).
>> Tested on ARMv5 by Andrew Lunn (andrew@lunn.ch).
>> Expected to work on ARMv6 as well, as its a part ARMv7 and part ARMv5.
>> Although, a proper testing is not done for ARMv6.
>>
>> Both of these testing are done with and without CONFIG_FRAME_POINTER
>> separately for LITTLE ENDIAN machine.
>>
>> For testing:
>>
>> 1. JIT is enabled with
>>          echo 1 > /proc/sys/net/core/bpf_jit_enable
>> 2. Constant Blinding can be enabled along with JIT using
>>          echo 1 > /proc/sys/net/core/bpf_jit_enable
>>          echo 2 > /proc/sys/net/core/bpf_jit_harden
>>
>> See Documentation/networking/filter.txt for more information.
>>
>> Result : test_bpf: Summary: 314 PASSED, 0 FAILED, [278/306 JIT'ed]

Did you also manage to get the BPF selftest suite running in the meantime
(tools/testing/selftests/bpf/)? There are a couple of programs that clang
will compile (test_pkt_access.o, test_xdp.o, test_l4lb.o, test_tcp_estats.o)
and then test run.

Did you manage to get tail calls tested as well (I assume so since you
implemented emit_bpf_tail_call() in the patch but just out of curiosity)?

>> Signed-off-by: Shubham Bansal <illusionist.neo@gmail.com>
>> ---
>>   Documentation/networking/filter.txt |    4 +-
>>   arch/arm/Kconfig                    |    2 +-
>>   arch/arm/net/bpf_jit_32.c           | 2404 ++++++++++++++++++++++++-----------
>>   arch/arm/net/bpf_jit_32.h           |  108 +-
>>   4 files changed, 1713 insertions(+), 805 deletions(-)
>>
[...]

If arm folks take the patch, there will be two minor (silent) merge
conflicts with net-next:

1) In bpf_int_jit_compile(), below the jited = 1 assignment, there
    needs to come a prog->jited_len = image_size.
2) The internal tail call opcode changed from BPF_JMP | BPF_CALL | BPF_X
    into BPF_JMP | BPF_TAIL_CALL.

Two minor things below, could probably also be as follow-up.

[...]
>> +       /* dst = imm64 */
>> +       case BPF_LD | BPF_IMM | BPF_DW:
>> +       {
>> +               const struct bpf_insn insn1 = insn[1];
>> +               u32 hi, lo = imm;
>> +
>> +               if (insn1.code != 0 || insn1.src_reg != 0 ||
>> +                   insn1.dst_reg != 0 || insn1.off != 0) {
>> +                       /* Note: verifier in BPF core must catch invalid
>> +                        * instruction.
>> +                        */
>> +                       pr_err_once("Invalid BPF_LD_IMM64 instruction\n");
>> +                       return -EINVAL;
>> +               }

Nit: this check can be removed as verifier already takes care
of it. (No JIT checks for this anymore.)

>> +               hi = insn1.imm;
>> +               emit_a32_mov_i(dst_lo, lo, dstk, ctx);
>> +               emit_a32_mov_i(dst_hi, hi, dstk, ctx);
>> +
>> +               return 1;
>> +       }
[...]
>> -       /* compute offsets only during the first pass */
>> -       if (ctx->target == NULL)
>> -               ctx->offsets[i] = ctx->idx * 4;
>> +static int validate_code(struct jit_ctx *ctx)
>> +{
>> +       int i;
>> +
>> +       for (i = 0; i < ctx->idx; i++) {
>> +               u32 a32_insn = le32_to_cpu(ctx->target[i]);

Given __opcode_to_mem_arm(ARM_INST_UDF) is used to fill the image,
perhaps use the __mem_to_opcode_arm() helper for the check?

>> +               if (a32_insn == ARM_INST_UDF)
>> +                       return -1;
>> +       }
>>
>>          return 0;
>>   }
>>
>> +void bpf_jit_compile(struct bpf_prog *prog)
>> +{
>> +       /* Nothing to do here. We support Internal BPF. */
>> +}

[toc] | [next] | [standalone]

#1663424

On Mon, Jun 12, 2017 at 12:21:03PM +0200, Daniel Borkmann wrote:
> On 05/30/2017 09:19 PM, Kees Cook wrote:
> >On Thu, May 25, 2017 at 4:13 PM, Shubham Bansal
> ><illusionist.neo@gmail.com> wrote:
> >>+static int validate_code(struct jit_ctx *ctx)
> >>+{
> >>+       int i;
> >>+
> >>+       for (i = 0; i < ctx->idx; i++) {
> >>+               u32 a32_insn = le32_to_cpu(ctx->target[i]);
> 
> Given __opcode_to_mem_arm(ARM_INST_UDF) is used to fill the image,
> perhaps use the __mem_to_opcode_arm() helper for the check?
> 
> >>+               if (a32_insn == ARM_INST_UDF)

The following is probably better:

		if (ctx->target[i] == __opcode_to_mem_arm(ARM_INST_UDF))

since then you can take advantage of the compiler optimising the
constant rather than having to do a byte swap on an unknown 32-bit
value.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

[toc] | [prev] | [next] | [standalone]

#1663730

Hi Russel,

On Mon, Jun 12, 2017 at 4:36 PM, Russell King - ARM Linux
<linux@armlinux.org.uk> wrote:
> On Mon, Jun 12, 2017 at 12:21:03PM +0200, Daniel Borkmann wrote:
>> On 05/30/2017 09:19 PM, Kees Cook wrote:
>> >On Thu, May 25, 2017 at 4:13 PM, Shubham Bansal
>> ><illusionist.neo@gmail.com> wrote:
>> >>+static int validate_code(struct jit_ctx *ctx)
>> >>+{
>> >>+       int i;
>> >>+
>> >>+       for (i = 0; i < ctx->idx; i++) {
>> >>+               u32 a32_insn = le32_to_cpu(ctx->target[i]);
>>
>> Given __opcode_to_mem_arm(ARM_INST_UDF) is used to fill the image,
>> perhaps use the __mem_to_opcode_arm() helper for the check?
>>
>> >>+               if (a32_insn == ARM_INST_UDF)
>
> The following is probably better:
>
>                 if (ctx->target[i] == __opcode_to_mem_arm(ARM_INST_UDF))
>
> since then you can take advantage of the compiler optimising the
> constant rather than having to do a byte swap on an unknown 32-bit
> value.

Done. Thanks :)
Please check if you can find anymore issues with the code. I really
appreciate it.

-Shubham

[toc] | [prev] | [next] | [standalone]

#1663752

On Mon, Jun 12, 2017 at 3:51 PM, Daniel Borkmann <daniel@iogearbox.net> wrote:
> On 05/30/2017 09:19 PM, Kees Cook wrote:

>>> This patch is essentially changing the current implementation of JIT
>>> compiler of Berkeley Packet Filter from classic to internal with almost
>>> all instructions from eBPF ISA supported except the following
>>>          BPF_ALU64 | BPF_DIV | BPF_K
>>>          BPF_ALU64 | BPF_DIV | BPF_X
>>>          BPF_ALU64 | BPF_MOD | BPF_K
>>>          BPF_ALU64 | BPF_MOD | BPF_X
>>>          BPF_STX | BPF_XADD | BPF_W
>>>          BPF_STX | BPF_XADD | BPF_DW
>>>          BPF_JMP | BPF_CALL
>
>
> Any plans to implement above especially BPF_JMP | BPF_CALL in near future?
> Reason why I'm asking is that i) currently the arm32 cBPF JIT implements
> all of the cBPF extensions (except SKF_AD_RANDOM and SKF_AD_VLAN_TPID).
> Some of the programs that were JITed before e.g. using SKF_AD_CPU would now
> fall back to the eBPF interpreter due to lack of translation in JIT, but
> also ii) that probably most (if not all) of eBPF programs use BPF helper
> calls heavily, which will still redirect them to the interpreter right now
> due to lack of BPF_JMP | BPF_CALL support, so it's really quite essential
> to have it implemented.

I can try for BPF_JMP | BPF_CALL. I didn't do it last time because I
thought, it would make the code look messy and become pain to get it
through the review.
For this, I have to map eBPF arguments with arm ABI arguments and move
ebpf arguments to corresponding arm ABI arguments, as eBPF arguments
doesn't match with arm ABI arguments.
Let me try that if its possible.

As far as following 4 are concerned :

>>>          BPF_ALU64 | BPF_DIV | BPF_K
>>>          BPF_ALU64 | BPF_DIV | BPF_X
>>>          BPF_ALU64 | BPF_MOD | BPF_K
>>>          BPF_ALU64 | BPF_MOD | BPF_X

I don't think it possible with current constraints over registers. I
already tried this.

>
>>> Implementation is using scratch space to emulate 64 bit eBPF ISA on 32
>>> bit
>>> ARM because of deficiency of general purpose registers on ARM. Currently,
>>> only LITTLE ENDIAN machines are supported in this eBPF JIT Compiler.
>>>
>>> Tested on ARMv7 with QEMU by me (Shubham Bansal).
>>> Tested on ARMv5 by Andrew Lunn (andrew@lunn.ch).
>>> Expected to work on ARMv6 as well, as its a part ARMv7 and part ARMv5.
>>> Although, a proper testing is not done for ARMv6.
>>>
>>> Both of these testing are done with and without CONFIG_FRAME_POINTER
>>> separately for LITTLE ENDIAN machine.
>>>
>>> For testing:
>>>
>>> 1. JIT is enabled with
>>>          echo 1 > /proc/sys/net/core/bpf_jit_enable
>>> 2. Constant Blinding can be enabled along with JIT using
>>>          echo 1 > /proc/sys/net/core/bpf_jit_enable
>>>          echo 2 > /proc/sys/net/core/bpf_jit_harden
>>>
>>> See Documentation/networking/filter.txt for more information.
>>>
>>> Result : test_bpf: Summary: 314 PASSED, 0 FAILED, [278/306 JIT'ed]
>
>
> Did you also manage to get the BPF selftest suite running in the meantime
> (tools/testing/selftests/bpf/)? There are a couple of programs that clang
> will compile (test_pkt_access.o, test_xdp.o, test_l4lb.o, test_tcp_estats.o)
> and then test run.

Nope. It looks like a latest addition to testing. Can you please tell
me how to test with it?

>
> Did you manage to get tail calls tested as well (I assume so since you
> implemented emit_bpf_tail_call() in the patch but just out of curiosity)?

I didn't try it exclusively, I thought test_bpf must have tested it. Doesn't it?

>
>>> Signed-off-by: Shubham Bansal <illusionist.neo@gmail.com>
>>> ---
>>>   Documentation/networking/filter.txt |    4 +-
>>>   arch/arm/Kconfig                    |    2 +-
>>>   arch/arm/net/bpf_jit_32.c           | 2404
>>> ++++++++++++++++++++++++-----------
>>>   arch/arm/net/bpf_jit_32.h           |  108 +-
>>>   4 files changed, 1713 insertions(+), 805 deletions(-)
>>>
> [...]
>
> If arm folks take the patch, there will be two minor (silent) merge
> conflicts with net-next:
>
> 1) In bpf_int_jit_compile(), below the jited = 1 assignment, there
>    needs to come a prog->jited_len = image_size.

Done.

> 2) The internal tail call opcode changed from BPF_JMP | BPF_CALL | BPF_X
>    into BPF_JMP | BPF_TAIL_CALL.

Done.

>
> Two minor things below, could probably also be as follow-up.
>
> [...]
>>>
>>> +       /* dst = imm64 */
>>> +       case BPF_LD | BPF_IMM | BPF_DW:
>>> +       {
>>> +               const struct bpf_insn insn1 = insn[1];
>>> +               u32 hi, lo = imm;
>>> +
>>> +               if (insn1.code != 0 || insn1.src_reg != 0 ||
>>> +                   insn1.dst_reg != 0 || insn1.off != 0) {
>>> +                       /* Note: verifier in BPF core must catch invalid
>>> +                        * instruction.
>>> +                        */
>>> +                       pr_err_once("Invalid BPF_LD_IMM64
>>> instruction\n");
>>> +                       return -EINVAL;
>>> +               }
>
>
> Nit: this check can be removed as verifier already takes care
> of it. (No JIT checks for this anymore.)
>
>>> +               hi = insn1.imm;
>>> +               emit_a32_mov_i(dst_lo, lo, dstk, ctx);
>>> +               emit_a32_mov_i(dst_hi, hi, dstk, ctx);
>>> +
>>> +               return 1;
>>> +       }
>
> [...]
>>>
>>> -       /* compute offsets only during the first pass */
>>> -       if (ctx->target == NULL)
>>> -               ctx->offsets[i] = ctx->idx * 4;
>>> +static int validate_code(struct jit_ctx *ctx)
>>> +{
>>> +       int i;
>>> +
>>> +       for (i = 0; i < ctx->idx; i++) {
>>> +               u32 a32_insn = le32_to_cpu(ctx->target[i]);
>
>
> Given __opcode_to_mem_arm(ARM_INST_UDF) is used to fill the image,
> perhaps use the __mem_to_opcode_arm() helper for the check?

Done.


I will send the patch again with these fixes. I really appreciate if
you could find more issues with the code, so that I can add it to the
next fix.

Thanks.
Shubham

[toc] | [prev] | [next] | [standalone]

#1664278

On Mon, Jun 12, 2017 at 09:10:07PM +0530, Shubham Bansal wrote:
> 
> Nope. It looks like a latest addition to testing. Can you please tell
> me how to test with it?
>
cd tools/testing/selftests/bpf/
make
sudo ./test_progs

-- 
Mit freundlichen Grüßen

Alexander Alemayhu

[toc] | [prev] | [next] | [standalone]

#1664281

From: Alexander Alemayhu <alexander@alemayhu.com>
Date: Tue, 13 Jun 2017 00:45:45 +0200

> On Mon, Jun 12, 2017 at 09:10:07PM +0530, Shubham Bansal wrote:
>> 
>> Nope. It looks like a latest addition to testing. Can you please tell
>> me how to test with it?
>>
> cd tools/testing/selftests/bpf/
> make
> sudo ./test_progs

Also, you might need to do a "make headers_install" at the top level
before doing this.

[toc] | [prev] | [next] | [standalone]

#1664292

On 06/12/2017 05:40 PM, Shubham Bansal wrote:
[...]
>> Did you manage to get tail calls tested as well (I assume so since you
>> implemented emit_bpf_tail_call() in the patch but just out of curiosity)?
>
> I didn't try it exclusively, I thought test_bpf must have tested it. Doesn't it?

In samples/bpf/ there's sockex3* that would exercise it, or
alternatively in iproute2 repo under examples/bpf/ there's
bpf_cyclic.c and bpf_tailcall.c as a prog.

Hm, generally, we should really add a test case also to BPF
selftest suite to facilitate that. I'll likely do that for
the next batch of BPF patches.

[toc] | [prev] | [next] | [standalone]

#1664508

Hi Daniel, Kees, David, Russel,

>> Any plans to implement above especially BPF_JMP | BPF_CALL in near future?
>> Reason why I'm asking is that i) currently the arm32 cBPF JIT implements
>> all of the cBPF extensions (except SKF_AD_RANDOM and SKF_AD_VLAN_TPID).
>> Some of the programs that were JITed before e.g. using SKF_AD_CPU would now
>> fall back to the eBPF interpreter due to lack of translation in JIT, but
>> also ii) that probably most (if not all) of eBPF programs use BPF helper
>> calls heavily, which will still redirect them to the interpreter right now
>> due to lack of BPF_JMP | BPF_CALL support, so it's really quite essential
>> to have it implemented.
>
> I can try for BPF_JMP | BPF_CALL. I didn't do it last time because I
> thought, it would make the code look messy and become pain to get it
> through the review.
> For this, I have to map eBPF arguments with arm ABI arguments and move
> ebpf arguments to corresponding arm ABI arguments, as eBPF arguments
> doesn't match with arm ABI arguments.
> Let me try that if its possible.

Okay. I looked at it, tried few different solutions also. There is a
problem with implementing BPF_JMP | BPF_CALL.
Problem is transition between 4 byte and 8 byte arguments. Lets take a
look a the following example to get a more clear look at the problem.

Lets consider this function :
CASE 1:                            foo(int a, int b, long long c, int d)
For calling this function in arm 32 arch, I have to pass the arguments
as following:
                                         a -> r0
                                         b -> r1
                                         c -> r2, r3
                                         d -> stack_top

Now consider an another example function :
CASE 2:                           bar(int a, int b, int c, int d)
For calling this function in arm32 arch, I have to pass the arguments
as following:
                                       a -> r0
                                       b -> r1
                                       c -> r2
                                       d -> r3

So, you can clearly see the problem with it. There is no way of
knowing which of the above way to pass the arguments. There are
solutions possible:

1. One thing I can do is look at the address of the function to call
and pass the argument accordingly but thats not really a robust
solution as we have to change the arm32 JIT each time we add any new
BPF helper function.

2. Another solution is, if any of you guys can assure/confirm me that
there will be only 4 byte argument passed to BPF helper functions in
arm32 as of now and in future including the pointer as well, then I
can just assume that each argument is passed as 4 byte value and my
trimming the 8byte arguments to 4 bytes arguments wouldn't be a
problem. In that case, arguments for CASE 1 and CASE 2 will be passed
in the same way, i.e.
                                       a -> r0
                                       b -> r1
                                       c -> r2
                                       d -> r3

Let me know what you think. I don't think I can find the solution to
this problem other than those mentioned above. Would love to here any
ideas from you guys.

>> Did you also manage to get the BPF selftest suite running in the meantime
>> (tools/testing/selftests/bpf/)? There are a couple of programs that clang
>> will compile (test_pkt_access.o, test_xdp.o, test_l4lb.o, test_tcp_estats.o)
>> and then test run.
I will run these tests tonight. Hopefully I will be able to run them.

Any comments are welcome. Would love to here what you think about my
solutions above.

Thanks.
Shubham

[toc] | [prev] | [next] | [standalone]

#1666206

On 06/13/2017 08:56 AM, Shubham Bansal wrote:
> Hi Daniel, Kees, David, Russel,
>
>>> Any plans to implement above especially BPF_JMP | BPF_CALL in near future?
>>> Reason why I'm asking is that i) currently the arm32 cBPF JIT implements
>>> all of the cBPF extensions (except SKF_AD_RANDOM and SKF_AD_VLAN_TPID).
>>> Some of the programs that were JITed before e.g. using SKF_AD_CPU would now
>>> fall back to the eBPF interpreter due to lack of translation in JIT, but
>>> also ii) that probably most (if not all) of eBPF programs use BPF helper
>>> calls heavily, which will still redirect them to the interpreter right now
>>> due to lack of BPF_JMP | BPF_CALL support, so it's really quite essential
>>> to have it implemented.
>>
>> I can try for BPF_JMP | BPF_CALL. I didn't do it last time because I
>> thought, it would make the code look messy and become pain to get it
>> through the review.
>> For this, I have to map eBPF arguments with arm ABI arguments and move
>> ebpf arguments to corresponding arm ABI arguments, as eBPF arguments
>> doesn't match with arm ABI arguments.
>> Let me try that if its possible.
>
> Okay. I looked at it, tried few different solutions also. There is a
> problem with implementing BPF_JMP | BPF_CALL.
> Problem is transition between 4 byte and 8 byte arguments. Lets take a
> look a the following example to get a more clear look at the problem.
>
> Lets consider this function :
> CASE 1:                            foo(int a, int b, long long c, int d)
> For calling this function in arm 32 arch, I have to pass the arguments
> as following:
>                                           a -> r0
>                                           b -> r1
>                                           c -> r2, r3
>                                           d -> stack_top
>
> Now consider an another example function :
> CASE 2:                           bar(int a, int b, int c, int d)
> For calling this function in arm32 arch, I have to pass the arguments
> as following:
>                                         a -> r0
>                                         b -> r1
>                                         c -> r2
>                                         d -> r3
>
> So, you can clearly see the problem with it. There is no way of
> knowing which of the above way to pass the arguments. There are
> solutions possible:

Right.

> 1. One thing I can do is look at the address of the function to call
> and pass the argument accordingly but thats not really a robust
> solution as we have to change the arm32 JIT each time we add any new
> BPF helper function.

Yeah, that would be rather ugly.

> 2. Another solution is, if any of you guys can assure/confirm me that
> there will be only 4 byte argument passed to BPF helper functions in
> arm32 as of now and in future including the pointer as well, then I
> can just assume that each argument is passed as 4 byte value and my
> trimming the 8byte arguments to 4 bytes arguments wouldn't be a
> problem. In that case, arguments for CASE 1 and CASE 2 will be passed
> in the same way, i.e.
>                                         a -> r0
>                                         b -> r1
>                                         c -> r2
>                                         d -> r3
>
> Let me know what you think. I don't think I can find the solution to
> this problem other than those mentioned above. Would love to here any
> ideas from you guys.

Not all of the helpers have 4 or less byte arguments only, there are a
few with 8 byte arguments, so making that general assumption wouldn't
work. I guess what could be done is that helpers have a flag in struct
bpf_func_proto which indicates for JITs that all args are 4 byte on 32bit
so you could probably use convention similar to case2 for them. Presumably
for that information to process, the JIT might need to be reworked to
extract that via bpf_analyzer() that does a verifier run to re-analyze
the program like in nfp JIT case.

The other option could perhaps be to check the interpreter disasm of
___bpf_prog_run() with regards to how it handles BPF_JMP | BPF_CALL
helper call and do something similarly generic in the JIT as well.

>>> Did you also manage to get the BPF selftest suite running in the meantime
>>> (tools/testing/selftests/bpf/)? There are a couple of programs that clang
>>> will compile (test_pkt_access.o, test_xdp.o, test_l4lb.o, test_tcp_estats.o)
>>> and then test run.
> I will run these tests tonight. Hopefully I will be able to run them.

Ok.

> Any comments are welcome. Would love to here what you think about my
> solutions above.

[toc] | [prev] | [next] | [standalone]

#1668243

Hi Daniel,

>
> Not all of the helpers have 4 or less byte arguments only, there are a
> few with 8 byte arguments, so making that general assumption wouldn't
> work. I guess what could be done is that helpers have a flag in struct
> bpf_func_proto which indicates for JITs that all args are 4 byte on 32bit
> so you could probably use convention similar to case2 for them. Presumably
> for that information to process, the JIT might need to be reworked to
> extract that via bpf_analyzer() that does a verifier run to re-analyze
> the program like in nfp JIT case.

Let me try a better solution which can be used to support both 4 byte
and 8 byte arguments. I hope it would work out. Are you sure this
patch can pass if it only supports 4 byte arguments though?
Let me list out what I have to do, so that you can tell me if I am
thinking in a wrong way :-

* I will add a bit flag in bpf_func_proto to represent whether
different arguments in a function call are 4 bytes or 8 bytes. If lsb
of bit flag is set then first argument is 8 byte, otherwise its not. I
think I can handle this flag properly in build_insn() in my code. Does
this sound okay?

I don't understand second part of your solution, i.e.

> Presumably
> for that information to process, the JIT might need to be reworked to
> extract that via bpf_analyzer() that does a verifier run to re-analyze
> the program like in nfp JIT case.

Please explain what are you suggesting and how can I extract bit flag
from bpf_func_proto().

Please reply asap, as I would like to finish it over the weekend. Please.

-Shubham

[toc] | [prev] | [next] | [standalone]

#1669568

On 06/17/2017 02:23 PM, Shubham Bansal wrote:
> Hi Daniel,
>
>> Not all of the helpers have 4 or less byte arguments only, there are a
>> few with 8 byte arguments, so making that general assumption wouldn't
>> work. I guess what could be done is that helpers have a flag in struct
>> bpf_func_proto which indicates for JITs that all args are 4 byte on 32bit
>> so you could probably use convention similar to case2 for them. Presumably
>> for that information to process, the JIT might need to be reworked to
>> extract that via bpf_analyzer() that does a verifier run to re-analyze
>> the program like in nfp JIT case.
>
> Let me try a better solution which can be used to support both 4 byte
> and 8 byte arguments. I hope it would work out. Are you sure this
> patch can pass if it only supports 4 byte arguments though?
> Let me list out what I have to do, so that you can tell me if I am
> thinking in a wrong way :-
>
> * I will add a bit flag in bpf_func_proto to represent whether
> different arguments in a function call are 4 bytes or 8 bytes. If lsb
> of bit flag is set then first argument is 8 byte, otherwise its not. I
> think I can handle this flag properly in build_insn() in my code. Does
> this sound okay?
>
> I don't understand second part of your solution, i.e.
>
>> Presumably
>> for that information to process, the JIT might need to be reworked to
>> extract that via bpf_analyzer() that does a verifier run to re-analyze
>> the program like in nfp JIT case.
>
> Please explain what are you suggesting and how can I extract bit flag
> from bpf_func_proto().
>
> Please reply asap, as I would like to finish it over the weekend. Please.

Sorry, had a travel over the weekend, so didn't read it in time.

What is the issue with imitating in JIT what the interpreter is
doing as a starting point? That should be generic enough to handle
any case.

Otherwise you'd need some sort of reverse mapping since verifier
already converted BPF_CALL insns into relative helper addresses
in imm part.

> -Shubham
>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web