feat: enable lwip checksum in software as option#1267
Conversation
Dependency ReviewThe following issues were found:
Snapshot WarningsConsider enabling retry-on-snapshot-warnings. See the documentation for more information and troubleshooting advice. License Issueslwip/lwip/CMakeLists.txt
external/args/CMakeLists.txt
external/segger_rtt/CMakeLists.txt
external/protobuf/CMakeLists.txt
osal/freertos/CMakeLists.txt
osal/threadx/CMakeLists.txt
external/crypto/micro-ecc/CMakeLists.txt
external/crypto/mbedtls/CMakeLists.txt
external/crypto/tiny-aes128/CMakeLists.txt
infra/syntax/CMakeLists.txt
cmake/emil_test_helpers.cmake
OpenSSF ScorecardScorecard details
Scanned Files
|
There was a problem hiding this comment.
Pull request overview
This PR adds a build-time option to enable lwIP checksum generation/verification in software (for hardware/stacks without checksum offload) and fixes a potential TLS receive stall in ConnectionMbedTls when mbedTLS consumes encrypted bytes but returns WANT_READ.
Changes:
- Add
EMIL_LWIP_CHECKSUM_IN_SOFTWARECMake option and wire it to lwIP checksum-related config macros viaLWIP_CHECKSUM_IN_SOFTWARE. - Prevent TLS receive stalling by retrying
mbedtls_ssl_read()when the SSL BIO actually provided data even if mbedTLS returnsMBEDTLS_ERR_SSL_WANT_READ. - Track whether the SSL receive callback provided any bytes during a read attempt.
Reviewed changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| services/network/ConnectionMbedTls.hpp | Adds state (sslReceiveProvidedData) to track whether the mbedTLS receive callback returned any data. |
| services/network/ConnectionMbedTls.cpp | Uses sslReceiveProvidedData to decide when to continue looping on MBEDTLS_ERR_SSL_WANT_READ, avoiding a potential stall. |
| lwip/lwip_config/lwipopts.h | Introduces LWIP_CHECKSUM_IN_SOFTWARE and ties lwIP checksum macros to it. |
| lwip/lwip_config/CMakeLists.txt | Applies LWIP_CHECKSUM_IN_SOFTWARE=1 when EMIL_LWIP_CHECKSUM_IN_SOFTWARE is enabled. |
| CMakeLists.txt | Adds the EMIL_LWIP_CHECKSUM_IN_SOFTWARE configuration option. |
✅
|
| Descriptor | Linter | Files | Fixed | Errors | Warnings | Elapsed time |
|---|---|---|---|---|---|---|
| ✅ ACTION | actionlint | 11 | 0 | 0 | 0.3s | |
| ✅ ACTION | zizmor | 11 | 0 | 0 | 0 | 3.1s |
| ✅ CPP | clang-format | 1096 | 8 | 0 | 0 | 10.84s |
| ✅ DOCKERFILE | hadolint | 2 | 0 | 0 | 0.05s | |
| ✅ JSON | jsonlint | 7 | 0 | 0 | 0.33s | |
| ✅ JSON | prettier | 7 | 0 | 0 | 0 | 0.61s |
| markdownlint | 8 | 0 | 5 | 0 | 1.38s | |
| ✅ MARKDOWN | markdown-table-formatter | 8 | 0 | 0 | 0 | 0.37s |
| ✅ REPOSITORY | checkov | yes | no | no | 36.48s | |
| ✅ REPOSITORY | git_diff | yes | no | no | 0.09s | |
| ✅ REPOSITORY | grype | yes | no | no | 56.91s | |
| ✅ REPOSITORY | ls-lint | yes | no | no | 0.01s | |
| osv-scanner | yes | 1 | no | 0.67s | ||
| ✅ REPOSITORY | secretlint | yes | no | no | 18.26s | |
| ✅ REPOSITORY | syft | yes | no | no | 2.14s | |
| ✅ REPOSITORY | trivy | yes | no | no | 20.18s | |
| ✅ REPOSITORY | trivy-sbom | yes | no | no | 0.26s | |
| ✅ REPOSITORY | trufflehog | yes | no | no | 14.63s | |
| lychee | 140 | 1 | 0 | 116.61s | ||
| prettier | 21 | 1 | 1 | 0 | 0.7s | |
| ✅ YAML | v8r | 21 | 0 | 0 | 10.44s | |
| ✅ YAML | yamllint | 21 | 0 | 0 | 0.74s |
Detailed Issues
⚠️ SPELL / lychee - 1 error
📝 Summary
---------------------
🔍 Total..........697
🔗 Unique.........661
✅ Successful.....691
⏳ Timeouts.........0
🔀 Redirected.....310
👻 Excluded.........5
❓ Unknown..........0
🚫 Errors...........1
⛔ Unsupported......1
Errors in external/protoc/CMakeLists.txt
[404] https://github.com/protocolbuffers/protobuf/releases/download/v$%7Bprotobuf_tag%7D/protoc-$%7Bprotobuf_version%7D-$%7Bos_postfix%7D.zip (at 18:13) | Rejected status code: 404 Not Found
Hint: Followed 310 redirects. You might want to consider replacing redirecting URLs with the resolved URLs. Use verbose mode (`-v`/`-vv`) to see redirection details.
Hint: You can configure accepted/rejected response codes with `-a` or `--accept`
⚠️ MARKDOWN / markdownlint - 5 errors
.github/instructions/microtest.instructions.md:7 error MD041/first-line-heading/first-line-h1 First line in a file should be a top-level heading [Context: "## Google Test Suite Coding Ru..."]
external/crypto/tiny-aes128/README.md:1 error MD041/first-line-heading/first-line-h1 First line in a file should be a top-level heading [Context: "### Tiny AES128 in C"]
external/crypto/tiny-aes128/README.md:29 error MD046/code-block-style Code block style [Expected: fenced; Actual: indented]
external/crypto/tiny-aes128/README.md:39 error MD046/code-block-style Code block style [Expected: fenced; Actual: indented]
external/crypto/tiny-aes128/README.md:49 error MD046/code-block-style Code block style [Expected: fenced; Actual: indented]
⚠️ REPOSITORY / osv-scanner - 1 error
Scanning dir .
Starting filesystem walk for root: /
Scanned external/segger_rtt file and found 0 packages
Scanned external/args file and found 0 packages
Scanned external/protobuf file and found 0 packages
Scanned external/protoc file and found 0 packages
Scanned external/crypto file and found 0 packages
End status: 155 dirs visited, 1617 inodes visited, 5 Extract calls, 533.10055ms elapsed, 533.100851ms wall time
No package sources found, --help for usage information.
⚠️ YAML / prettier - 1 error
[error] Explicitly specified pattern "documents/modules/ROOT/examples/clangformat.yaml" is a symbolic link.
.clusterfuzzlite/project.yaml 30ms (unchanged)
.github/dependabot.yml 19ms (unchanged)
.github/workflows/ci.yml 74ms (unchanged)
.github/workflows/dependency-scanner.yml 9ms (unchanged)
.github/workflows/documentation.yml 13ms (unchanged)
.github/workflows/fuzzing-batch.yml 5ms (unchanged)
.github/workflows/fuzzing-cron.yml 8ms (unchanged)
.github/workflows/fuzzing-pr.yml 7ms (unchanged)
.github/workflows/linting-formatting.yml 16ms (unchanged)
.github/workflows/release-please.yml 14ms (unchanged)
.github/workflows/security.yml 5ms (unchanged)
.github/workflows/static-analysis.yml 11ms (unchanged)
.github/workflows/validate-pr.yml 27ms (unchanged)
.ls-lint.yml 3ms
.mega-linter.yml 3ms (unchanged)
antora-playbook-branch.yml 4ms (unchanged)
antora-playbook-site.yml 3ms (unchanged)
documents/antora.yml 3ms (unchanged)
documents/supplemental-ui/ui.yml 1ms (unchanged)
mull.yml 2ms (unchanged)
Notices
📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)
See detailed reports in MegaLinter artifacts
Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)
- Documentation: Custom Flavors
- Command:
npx mega-linter-runner@9.5.0 --custom-flavor-setup --custom-flavor-linters ACTION_ACTIONLINT,ACTION_ZIZMOR,CPP_CLANG_FORMAT,DOCKERFILE_HADOLINT,JSON_JSONLINT,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GRYPE,REPOSITORY_LS_LINT,REPOSITORY_OSV_SCANNER,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,SPELL_LYCHEE,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository
|



feat: enable lwip checksum in software as option
Some network stacks/hardware does not compute the checksum, which requires lwip do it by software
fix: communication stall after TLS session ticket.
mbedTLS can consume encrypted bytes from the receive buffer yet still return WANT_READ because it doesn't have a complete TLS record to yield application data yet. Previously the loop would unconditionally break and wait for the next DataReceived() notification. But since the underlying data was already consumed/acked in that pass, no new notification necessarily arrives — so buffered encrypted data could sit unprocessed and the connection would stall/deadlock.
By tracking whether the BIO provided data and looping again when it did, the code keeps draining available encrypted data instead of giving up prematurely, preventing the stall.