src: find .text section using dl_iterate_phdr

[gabrielschulhof]

Use dl_iterate_phdr(3) to find the mapping containing
__node_text_start instead of parsing /proc/self/maps.

Signed-off-by: @gabrielschulhof

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29798/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29799/

[bnoordhuis]

LGTM % some minor (mostly style) issues.

[bnoordhuis]

Just checking, is end inclusive or exclusive?

Aside: it looks like there's no real need for the dl_params->reference_sym field. This also works, wouldn't it?

auto reference_sym = reinterpret_cast<uintptr_t>(&__node_text_start);
if (reference_sym >= start && reference_sym <= end) {
// ...
}

[gabrielschulhof]

@bnoordhuis the reference_sym field is being used both inside the dl_iterate_phdr callback and outside of it. I thought it best not to assign reinterpret_cast<uintptr_t>(&__node_text_start); to two different variables because, if the symbol changes in the future, they might get out of sync.

[bnoordhuis]

For legibility, it'd be better to hoist the lambda out of the function call:

auto callback = [](dl_phdr_info* info, size_t, void* data) -> int {
// ...
};
if (1 == dl_iterate_phdr(callback, &dl_params)) {
// ...
}

[gabrielschulhof]

@bnoordhuis I added some more sanity checks to ensure that we create a region that has proper offsets and > 0 page count.

[gabrielschulhof]

... or rather, that we consider the region to have been found only if it has proper offsets and >0 page count.

[gabrielschulhof]

@bnoordhuis I addressed your review comments and added some sanity checks wrt. your comment regarding inclusive/exclusive comparison.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29817/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29818/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29819/

[gabrielschulhof]

Rebased in the hope that if fixes the "node ASAN" test.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29820/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/29827/

[gabrielschulhof]

@nodejs/collaborators sorry about the spam! Can this land without the "node ASAN" test? AFAICT it's failing for a lot of PRs.

[gabrielschulhof]

https://github.com/nodejs/node/actions?query=workflow%3A%22node+ASAN%22

[addaleax]

sorry about the spam! Can this land without the "node ASAN" test? AFAICT it's failing for a lot of PRs.

Yes. 👍

[gengjiawen]

@nodejs/collaborators sorry about the spam! Can this land without the "node ASAN" test? AFAICT it's failing for a lot of PRs.

We dealt with Node ASAN fail in #32116. Not sure we should cherrypick the commit to master, since the PR is pretty ready to launch.

Also +1 for this PR to launch.

[gabrielschulhof]

Landed in 25cb855.

[targos]

Depends on the large pages change to land on v12.x