From 5e51e14f449418c5340478732d0c8f38eaf80666 Mon Sep 17 00:00:00 2001 From: Samuel Kalbfleisch Date: Fri, 13 May 2022 05:55:12 +0200 Subject: [PATCH] Fixes --- README.md | 8 ++++++-- artifact-evaluation/README.md | 9 ++++++--- .../walkthrough_performance-benchmarks.md | 4 +++- 3 files changed, 15 insertions(+), 6 deletions(-) diff --git a/README.md b/README.md index 8ee40c1..99ef574 100644 --- a/README.md +++ b/README.md @@ -32,7 +32,10 @@ installed. See "Artifact Evaluation" below. ```sh # install dependencies # on Fedora: -sudo dnf install python3-pip python3-mypy python3-capstone glibc-static elfutils-libelf-devel dtc capstone-devel libdwarf-devel glib2-devel pixman-devel protobuf-devel protobuf-c-devel curl-devel jsoncpp-devel chrpath datamash telnet +sudo dnf install python3-pip python3-mypy python3-capstone glibc-static \ + elfutils-libelf-devel dtc capstone-devel libdwarf-devel glib2-devel \ + pixman-devel protobuf-devel protobuf-c-devel curl-devel jsoncpp-devel \ + chrpath datamash telnet pip install yq sortedcontainers # Rust via rustup (see https://rustup.rs) @@ -50,7 +53,8 @@ fs-testing/linux/build-kernel.sh nova git clone https://github.com/linux-pmfs/pmfs fs-testing/linux/pmfs # Note: Building PMFS requires gcc4, build from a suitable container. For example: -podman run --rm -v"$PWD/fs-testing/linux:/mnt" docker.io/library/gcc:4 sh -c 'apt-get update && apt-get install bc && /mnt/build-kernel.sh pmfs' +podman run --rm -v"$PWD/fs-testing/linux:/mnt" docker.io/library/gcc:4 \ + sh -c 'apt-get update && apt-get install bc && /mnt/build-kernel.sh pmfs' ``` diff --git a/artifact-evaluation/README.md b/artifact-evaluation/README.md index d2418b0..27e5391 100644 --- a/artifact-evaluation/README.md +++ b/artifact-evaluation/README.md @@ -12,7 +12,8 @@ its analysis in parallel, so make sure to provide plenty of memory (`-m`) and vCPUs (`-smp`). As a rough guideline, provide 2 GB of memory per vCPU. For example with QEMU/KVM: ``` -qemu-kvm -m 16G -smp 8 -display none -serial mon:stdio -device e1000,netdev=net0 -netdev user,id=net0,hostfwd=tcp::2222-:22 vinter.qcow2 +qemu-kvm -m 16G -smp 8 -display none -serial mon:stdio -device e1000,netdev=net0 \ + -netdev user,id=net0,hostfwd=tcp::2222-:22 vinter.qcow2 ``` Connect to the virtual machine via SSH. The password for the users vinter and @@ -35,8 +36,10 @@ fs-testing/scripts/run_getting-started.sh The script will put results into the directory `results_getting-started`. View a short summary of these results with the following commands: ``` -vinter_python/report-results.py analyze results_getting-started/vinter_python/vm_nova/getting-started/test_hello-world -vinter_python/report-results.py analyze results_getting-started/vinter_python/vm_pmfs/getting-started/test_hello-world +vinter_python/report-results.py analyze \ + results_getting-started/vinter_python/vm_nova/getting-started/test_hello-world +vinter_python/report-results.py analyze \ + results_getting-started/vinter_python/vm_pmfs/getting-started/test_hello-world ``` You can see that Vinter reports a violation of *single final state* for the diff --git a/artifact-evaluation/walkthrough_performance-benchmarks.md b/artifact-evaluation/walkthrough_performance-benchmarks.md index 8eb7f01..79c7a5f 100644 --- a/artifact-evaluation/walkthrough_performance-benchmarks.md +++ b/artifact-evaluation/walkthrough_performance-benchmarks.md @@ -12,7 +12,7 @@ fs-testing/scripts/script_run-paper-benchmark.sh The benchmark may take between 30 and 60 minutes to complete and writes its files to `results_paper-benchmark`. -Its final output should be similar to the following: +Its final output should be similar to the following (only placeholder values listed here): ``` == Figure 8 (pmemtrace.py) results == count mean sstdev @@ -39,5 +39,7 @@ and result in 7 unique semantic crash states. success ``` +NOTE: The benchmark results listed in the paper PDF are not yet up to date with the latest artifact version. The differences are not big though. + The tables should approximately reproduce the numbers in Figure 8 and Figure 9 (depending on the hardware used). The paragraph at the end of the output should further contain the same numbers as the corresponding sentence in ยง6.3.