Remove duplication, avoid handwritten assembly #1165
Conversation
✅ Deploy Preview for aya-rs-docs ready!Built without sensitive environment variables
To edit notification comments on pull requests, go to your Netlify site configuration. |
tamird
changed the title
|
cc @Billy99 I believe the last commit fixes a bug on big-endian. PTAL. |
|
I was incorrect, the FD always goes in the first instruction, so there was no bug here. |
tamird
force-pushed
the
dedupe
branch
3 times, most recently
from
fb807e7 to
f939005
Compare
aya/src/sys/bpf.rs
Outdated
|
|
||
| let gpl = b"GPL\0"; | ||
| let mov64_imm = (BPF_ALU64 | BPF_MOV | BPF_K).try_into().unwrap(); |
There was a problem hiding this comment.
Given we re-use this same program in many places, is there any benefit to making it const?
There was a problem hiding this comment.
Maybe? But the constants are u32 and the field is u8 - without an as cast there's no way to make it const IIUC.
| u.insn_cnt = insns.len() as u32; | ||
| u.insns = insns.as_ptr() as u64; | ||
| u.prog_type = bpf_prog_type::BPF_PROG_TYPE_SOCKET_FILTER as u32; | ||
| u.prog_type = bpf_prog_type::BPF_PROG_TYPE_TRACEPOINT as u32; |
There was a problem hiding this comment.
Is there a motivation to change the program type?
IIRC SOCKET_FILTER is the oldest of the BPF program types so should always be available.
There was a problem hiding this comment.
Yes, it's in the commit message.
Use
BPF_PROG_TYPE_TRACEPOINTinstead ofBPF_PROG_TYPE_SOCKET_FILTER
as the former seems to work with more feature detection functions.
is_perf_link_supported didn't work when this was SOCKET_FILTER.
aya/src/sys/bpf.rs
Outdated
|
|
||
| let gpl = b"GPL\0"; | ||
| u.license = gpl.as_ptr() as u64; | ||
| // TODO(tamird): remove when these are generated on the userspace side. |
There was a problem hiding this comment.
Since #1166 has merged, please let me know if you need me to press the button on the codegen workflow. You should be able to trigger it from Github Actions.
There was a problem hiding this comment.
Yep, working on that in #1168 since the constants I wanted weren't all generated.
| u.map_name[..name_len] | ||
| .copy_from_slice(unsafe { slice::from_raw_parts(name.as_ptr(), name_len) }); | ||
| let name_bytes = name.to_bytes_with_nul(); | ||
| let len = cmp::min(name_bytes.len(), u.map_name.len()); |
There was a problem hiding this comment.
Does this account for the nul byte required to terminate the string?
The string has to be 15 chars + 1 nul byte.
There was a problem hiding this comment.
Yep, note I'm using to_bytes_with_nul.
The safety requirements of this transmutation are simpler.
Use `BPF_PROG_TYPE_TRACEPOINT` instead of `BPF_PROG_TYPE_SOCKET_FILTER` as the former seems to work with more feature detection functions.
See individual commits.
This change is