Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add initial RTT configurability #2041

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

robamu
Copy link

@robamu robamu commented Aug 5, 2024

This adds configurability for the initial round-trip time, which has a use-case for something like space-applications where a higher round-trip time might be expected. This is a first draft PR, an end-to-end test with other QUIC stacks still needs to be done.

Copy link

github-actions bot commented Aug 5, 2024

Failed Interop Tests

QUIC Interop Runner, client vs. server

neqo-latest as client

neqo-latest as server

All results

Succeeded Interop Tests

QUIC Interop Runner, client vs. server

neqo-latest as client

neqo-latest as server

Unsupported Interop Tests

QUIC Interop Runner, client vs. server

neqo-latest as client

neqo-latest as server

Copy link

github-actions bot commented Aug 5, 2024

Firefox builds for this PR

The following builds are available for testing. Crossed-out builds did not succeed.

@@ -111,6 +111,10 @@ pub struct QuicParameters {
/// The idle timeout for connections, in seconds.
pub idle_timeout: u64,

#[arg(long = "init_rtt", default_value = "100")]
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
#[arg(long = "init_rtt", default_value = "100")]
#[arg(long = "init_rtt", default_value = DEFAULT_INITIAL_RTT)]

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can the macro take a Duration like that? I thought that it had to be a string. The other way to spell this is to use Option<u64> and use map_or to get the default.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, not sure that it can take a Duration. Was trying to suggest to avoid violating POLA and somehow use the same default here that the code is using.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd use Option and move the application of the default to the getter. That works, even if it means that the --help doesn't report the default.

pub const INITIAL_RTT: Duration = Duration::from_millis(100);
pub const DEFAULT_INITIAL_RTT: Duration = Duration::from_millis(100);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think you have to rename this.

@robamu robamu force-pushed the initial-rtt-configurability branch from ee6f44c to 5445216 Compare August 29, 2024 15:37
@larseggert
Copy link
Collaborator

@robamu it would be helpful it you could compile the PR and run the test suite locally, just to make sure that CI doesn't immediately fail.

@robamu
Copy link
Author

robamu commented Aug 29, 2024

Ah, I can do that.

@robamu
Copy link
Author

robamu commented Aug 29, 2024

I have problems running the tests currently..

neqo on  initial-rtt-configurability is 📦 v0.8.2 via 🦀 v1.80.1
❯ cargo test
   Compiling neqo-bin v0.8.2 (/home/rmueller/Rust/neqo/neqo-bin)
    Finished `test` profile [unoptimized + debuginfo] target(s) in 2.57s
     Running unittests src/lib.rs (target/debug/deps/neqo_bin-33111872ccada738)

running 1 test
0s  0ms ERROR Minimum NSS version of 3.103
 not supported
test tests::write_qlog_file ... FAILED

failures:

---- tests::write_qlog_file stdout ----
thread 'tests::write_qlog_file' panicked at neqo-bin/src/lib.rs:309:80:
called `Result::unwrap()` on an `Err` value: UnsupportedVersion
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    tests::write_qlog_file

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

error: test failed, to rerun pass `-p neqo-bin --lib`

I can check and build the project though and I have installed NSS v3.103 according to https://www.linuxfromscratch.org/blfs/view/svn/postlfs/nss.html .. is there something special I need to take care of when installing NSS from the sources for the neqo tests? I can not use the Ubuntu package provides NSS tools/libraries because the version is too old.

@larseggert
Copy link
Collaborator

Are you sure you are linking against the LFS version of NSS you installed and not some other version? Check with ldd?

I suggest you follow what's in the readme, i.e., check out the sources and set the environment variables so that neqo will do the right thing.

@robamu
Copy link
Author

robamu commented Aug 29, 2024

Running ldd yields the following:

neqo on  main [$] is 📦 v0.8.2 via 🦀 v1.80.1
❯ ldd /usr/lib/libnss3.so
/usr/lib/libnss3.so: /lib/x86_64-linux-gnu/libnssutil3.so: version `NSSUTIL_3.101' not found (required by /usr/lib/libnss3.so)
        linux-vdso.so.1 (0x00007ffce9f5d000)
        libnssutil3.so => /lib/x86_64-linux-gnu/libnssutil3.so (0x000072ffb3e2f000)
        libplc4.so => /lib/x86_64-linux-gnu/libplc4.so (0x000072ffb3e28000)
        libplds4.so => /lib/x86_64-linux-gnu/libplds4.so (0x000072ffb3e23000)
        libnspr4.so => /lib/x86_64-linux-gnu/libnspr4.so (0x000072ffb3de3000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x000072ffb3a00000)
        /lib64/ld-linux-x86-64.so.2 (0x000072ffb3fbd000)

so.. some problem with libnssutil3 oder some other missing library?

I also tried following the readme by setting NSS_DIR to the NSS sources path and then running a fresh build of neqo, but there seems to be some issue with the build.sh script of NSS.. I was only able to build NSS with make so far.

   Compiling neqo-crypto v0.8.2 (/home/rmueller/Rust/neqo/neqo-crypto)
error: failed to run custom build command for `neqo-crypto v0.8.2 (/home/rmueller/Rust/neqo/neqo-crypto)`

Caused by:
  process didn't exit successfully: `/home/rmueller/Rust/neqo/target/debug/build/neqo-crypto-5811bbeef5d53e68/build-script-build` (exit status: 101)
  --- stdout
  cargo:rustc-check-cfg=cfg(nss_nodb)
  cargo:rerun-if-env-changed=NSS_DIR
  NSPR [1/5] configure ...

  --- stderr
  ./build.sh: line 36: ../configure: No such file or directory
  thread 'main' panicked at neqo-crypto/build.rs:135:5:
  NSS build failed
  note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
warning: build failed, waiting for other jobs to finish...

I have gyp and ninja installed, so I am not sure what is going on here

@larseggert
Copy link
Collaborator

larseggert commented Aug 30, 2024

Running ldd yields the following:

neqo on  main [$] is 📦 v0.8.2 via 🦀 v1.80.1
❯ ldd /usr/lib/libnss3.so
/usr/lib/libnss3.so: /lib/x86_64-linux-gnu/libnssutil3.so: version `NSSUTIL_3.101' not found (required by /usr/lib/libnss3.so)

...

so.. some problem with libnssutil3 oder some other missing library?

This looks like at least some of your installation is NSS 3.101 and not 3.103, which would explain the error when you try and run it. If you want to use the LFS-installed NSS, you need to make sure it's installed correctly in the system paths, and that no other (older) versions are lingering.

I also tried following the readme by setting NSS_DIR to the NSS sources path and then running a fresh build of neqo, but there seems to be some issue with the build.sh script of NSS.. I was only able to build NSS with make so far.

   Compiling neqo-crypto v0.8.2 (/home/rmueller/Rust/neqo/neqo-crypto)

...

./build.sh: line 36: ../configure: No such file or directory
thread 'main' panicked at neqo-crypto/build.rs:135:5:
NSS build failed

This looks like the environment is not set up correctly. You need to check out NSS and NSPR into a directory hierarchy like this:

/your/absolute/path
nss
├── nss
└── nspr

Your environment then should be:

export NSS_DIR=/your/absolute/path/nss/nss
export NSS_TARGET=Release
export LD_LIBRARY_PATH $NSS_DIR/../dist/$NSS_TARGET/lib

@robamu
Copy link
Author

robamu commented Aug 30, 2024

I have managed to build and run the tests now with cargo nextest r , one test fails for me for both main and the PR branch:

--- STDOUT:              neqo-common hrtime::test::update_multi ---

running 1 test
sleep(1ms) → 1.061399ms Δ61.399µs
sleep(1ms) → 1.058584ms Δ58.584µs
sleep(2ms) → 2.048389ms Δ48.389µs
sleep(2ms) → 2.066653ms Δ66.653µs
sleep(3ms) → 3.074642ms Δ74.642µs
sleep(3ms) → 3.073069ms Δ73.069µs
sleep(5ms) → 5.066455ms Δ66.455µs
sleep(5ms) → 5.072656ms Δ72.656µs
sleep(8ms) → 8.064904ms Δ64.904µs
sleep(8ms) → 8.067559ms Δ67.559µs
sleep(10ms) → 10.076534ms Δ76.534µs
sleep(10ms) → 10.071084ms Δ71.084µs
sleep(12ms) → 12.064259ms Δ64.259µs
sleep(12ms) → 12.063176ms Δ63.176µs
sleep(15ms) → 15.066245ms Δ66.245µs
sleep(15ms) → 15.062899ms Δ62.899µs
sleep(20ms) → 20.067497ms Δ67.497µs
sleep(20ms) → 20.06357ms Δ63.57µs
sleep(25ms) → 25.062007ms Δ62.007µs
sleep(25ms) → 25.068479ms Δ68.479µs
sleep(30ms) → 30.326814ms Δ326.814µs
sleep(1ms) → 1.077138ms Δ77.138µs
sleep(30ms) → 31.640385ms Δ1.640385ms
sleep(1ms) → 1.551488ms Δ551.488µs
sleep(2ms) → 2.430786ms Δ430.786µs
sleep(2ms) → 3.138001ms Δ1.138001ms
sleep(3ms) → 3.478609ms Δ478.609µs
sleep(3ms) → 5.995656ms Δ2.995656ms
sleep(5ms) → 5.087443ms Δ87.443µs
sleep(8ms) → 8.070364ms Δ70.364µs
sleep(10ms) → 10.086934ms Δ86.934µs
sleep(12ms) → 12.115404ms Δ115.404µs
sleep(15ms) → 15.082845ms Δ82.845µs
sleep(1ms) → 1.072259ms Δ72.259µs
sleep(2ms) → 2.108822ms Δ108.822µs
sleep(3ms) → 3.101262ms Δ101.262µs
sleep(5ms) → 6.947169ms Δ1.947169ms
sleep(20ms) → 20.086693ms Δ86.693µs
sleep(25ms) → 25.072748ms Δ72.748µs
sleep(30ms) → 30.073519ms Δ73.519µs
test hrtime::test::update_multi ... FAILED

failures:

failures:
    hrtime::test::update_multi

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 64 filtered out; finished in 0.27s


--- STDERR:              neqo-common hrtime::test::update_multi ---
thread '<unnamed>' panicked at neqo-common/src/hrtime.rs:416:38:
called `Result::unwrap()` on an `Err` value: ()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
thread 'hrtime::test::update_multi' panicked at neqo-common/src/hrtime.rs:475:20:
called `Result::unwrap()` on an `Err` value: Any { .. }

I see that there is a client & server test and the interoperability tests, should I try to make those work locally as well?

@larseggert
Copy link
Collaborator

Are you running this in a VM or cloud instance? Because those have bad timer accuracy. IIRC there is either an environment variable or rust feature you can use to skip those timer tests.

@larseggert
Copy link
Collaborator

Also, if cargo nextest run --no-fail-fast otherwise passes, i.e., if the timer test failure is the only failure, please update the PR and I'll have CI run.

Copy link

codecov bot commented Sep 13, 2024

Codecov Report

Attention: Patch coverage is 62.50000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 95.32%. Comparing base (7e8d22d) to head (1b1ce85).
Report is 18 commits behind head on main.

Files with missing lines Patch % Lines
neqo-transport/src/connection/params.rs 62.50% 3 Missing ⚠️
Additional details and impacted files
@@            Coverage Diff             @@
##             main    #2041      +/-   ##
==========================================
- Coverage   95.36%   95.32%   -0.04%     
==========================================
  Files         112      112              
  Lines       36505    36279     -226     
==========================================
- Hits        34813    34584     -229     
- Misses       1692     1695       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@robamu
Copy link
Author

robamu commented Sep 13, 2024

I actually ran this on my (decent) machine.. I can try to have another look to continue this on the weekend.. Maybe some more unittests would be good as well?

@larseggert
Copy link
Collaborator

Yes, in order for us to land this, CI needs to pass and test coverage needs to be there. (And we need some review approvals.)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants