tree 9dedcbf92098e6f251fd8c52955cbb466b0fb240
parent eff585190aca2ba62872eae3623667fda95f4d36
parent 9b7a489f451a54fab6f6534a3b4a48c532726dcb
author bors[bot] <26634292+bors[bot]@users.noreply.github.com> 1612912232 +0000
committer GitHub <noreply@github.com> 1612912232 +0000
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJgIxZoCRBK7hj4Ov3rIwAAdHIIAKLLJrpQjJG6wdzI4PAakYA6
 r9jG3jWjsa+zCkx1cR3UUTxtNYWLW62C/AWbnpQcTxpC6Tu0cC7ewvJG8YEAcpOY
 LQWyrcp3AZHUNUT4e9pkXxeNHgTjsOzEjw7q+Fmxn5ae0fuft1qg9lG0gT5M82fS
 JNe+zcA/Xwn/IvoxmCcvMryfcHjiGuP+buhEIpXuXc8EA2aYkCWsscdgaBptXCsG
 JQrVwScRz5/wYNxdwUaKC9HLc3oz2gtoP8UdlwEEzPxleGugBEn1WbtnVS9Jc+B2
 42Gr3FZsyTk6Q29Nz9Pu/WV8UQPq+ffu2qN7Cms/igv7agfLaVjCVJpNqmHtSZY=
 =z8B1
 -----END PGP SIGNATURE-----
 

Merge #272

272: Add the RISC-V entry point assembly to libtock_runtime r=hudson-ayers a=jrvanwhy

To support the entry point assembly, I rewrote the linker script ([link to the new version](https://github.com/jrvanwhy/libtock-rs/blob/riscv-asm/core/runtime/layout_generic.ld) in case you don't want to see the diff).

The rust toolchain is unable to compile raw assembly files on its own. I don't want a separate toolchain to be a build requirement of `libtock-rs`, so instead I committed the compiled library into the repository. To keep the compiled library and source code in sync, I added an actions workflow that verifies the compiled library matches the provided sources.

There is one part of this PR I am unhappy with. I have to manually `strip` local symbols from the compiled library -- i.e., removing the `riscv64-linux-gnu-strip` invocation from `assemble.sh` gives the following:

```
Disassembly of section .start:

00000000 <start>:
   0:   00000417                auipc   s0,0x0
   4:   87aa                    mv      a5,a0
   6:   4384                    lw      s1,0(a5)
   8:   00940c63                beq     s0,s1,20 <.Lset_brk>
   c:   4521                    li      a0,8
   e:   4585                    li      a1,1
  10:   4609                    li      a2,2
  12:   4709                    li      a4,2
  14:   00000073                ecall
  18:   4501                    li      a0,0
  1a:   4719                    li      a4,6
  1c:   00000073                ecall

00000020 <.Lset_brk>:
  20:   4501                    li      a0,0
  22:   43cc                    lw      a1,4(a5)
  24:   4715                    li      a4,5
  26:   00000073                ecall
  2a:   0087a103                lw      sp,8(a5)
  2e:   47c8                    lw      a0,12(a5)
  30:   c909                    beqz    a0,42 <.Lzero_bss>
  32:   4b8c                    lw      a1,16(a5)
  34:   4bd0                    lw      a2,20(a5)

00000036 <.Ldata_loop_body>:
  36:   4194                    lw      a3,0(a1)
  38:   c214                    sw      a3,0(a2)
  3a:   1571                    addi    a0,a0,-4
  3c:   0591                    addi    a1,a1,4
  3e:   0611                    addi    a2,a2,4
  40:   f97d                    bnez    a0,36 <.Ldata_loop_body>

00000042 <.Lzero_bss>:
  42:   4f88                    lw      a0,24(a5)
  44:   c519                    beqz    a0,52 <.Lcall_rust_start>
  46:   4fcc                    lw      a1,28(a5)

00000048 <.Lbss_loop_body>:
  48:   00058023                sb      zero,0(a1)
  4c:   157d                    addi    a0,a0,-1
  4e:   0585                    addi    a1,a1,1
  50:   fd65                    bnez    a0,48 <.Lbss_loop_body>

00000052 <.Lcall_rust_start>:
  52:   fafff0ef                jal     ra,0 <start>
```

I'd appreciate advice on how to prevent `riscv64-linux-gnu-as` from including the local symbols in its output.

Co-authored-by: Johnathan Van Why <jrvanwhy@google.com>