[EagleBearerYS]

Title-less conformance fixes of this kind are merged now, for example #7365 and #8411.

What

setsockopt reaches a socket created with sceNetSocket through the option setter libSceNet registers with libkernel, but getsockopt looks the descriptor up only among libkernel's own sockets and fails. A title can set an option on such a socket through the BSD call and not read it back.

• libSceNet now registers an option getter next to its setter, and libkernel's getsockopt asks it for a descriptor libkernel doesn't own, as setsockopt does. The getter answers through sceNetGetsockopt, so SO_RCVTIMEO and SO_SNDTIMEO come back as a struct timeval, as FreeBSD 11.4's sogetopt returns them.
• sceNetGetsockopt reads back SO_REUSEADDR and SO_BROADCAST, which sceNetSetsockopt takes but sceNetGetsockopt failed with EOPNOTSUPP. It reads them from the host socket. FreeBSD's sogetopt returns so_options & option for the boolean options, so they read back as 0x4 and 0x20 when set.
• The debt entry that lists the libkernel calls reaching a libSceNet socket now names getsockopt with setsockopt.

Two commits: e887616 for sceNetGetsockopt, aea804d for the libkernel getter and the debt entry. No title is claimed.

Overlap, merge order only: this PR goes after #7511 and #9365, and rebases on them. #7511 adds pointer and length checks at the start of getsockopt_nid_postfix, where this PR adds the getter call; after the rebase the call comes after those checks, so a libSceNet socket fails a bad pointer or length as libkernel's own sockets do. #9365 corrects the same close and setsockopt debt sentence, and this PR's addition of getsockopt to it goes on top. #9542 (Queuse) adds the same SO_REUSEADDR and SO_BROADCAST readback in sceNetGetsockopt, with short-buffer handling. The order is agreed with Queuse in this thread (comment): #9523 → #9542 → this PR → #9578. The two merge as text, but would put the same two case labels in one switch, so this PR drops e887616's sceNetGetsockopt hunk at its first merge of main after #9542 lands. It keeps the libkernel getter forwarding and its tests. On #9542 with this PR merged in and that hunk dropped, guest_sce_net and guest_net_boolean_readback pass on both runners (run).

Tested

On GitHub's runners from my fork, on main at a6dcea9 (run, of the same code; aea804d adds the debt edit only):

• guest_sce_net, on a datagram socket made with sceNetSocket:
  • sceNetGetsockopt reads SO_BROADCAST and SO_REUSEADDR back as 0x20 and 0x4 after sceNetSetsockopt sets them;
  • libkernel's getsockopt reads SO_BROADCAST back as 0x20;
  • after libkernel's setsockopt sets SO_RCVTIMEO to 0.25 s, libkernel's getsockopt returns the same struct timeval.
• With main's code guest_sce_net fails on Linux and Windows. With the change it passes 10 runs in a row on both.
• Every test registered under core/libs/prx/libc and core/libs/prx/libkernel passes: 142 and 207 on Windows, 197 libkernel tests on Linux. In the Linux libc suite (115), the timing test precise_wait failed with and without the change in this run.
• tools/check_conventions.py --base origin/main passes.
• xXPBSXx's run (comment): WSL2 Ubuntu 24.04 (a virtualised kernel), GCC 13.3.0, Release, the PR head alone on its merge base a6dcea9. guest_sce_net passes 3 of 3, and with libSceNet's Export.cpp and libkernel's SocketPoll.hpp, Socket.cpp and Udp.cpp at the merge base and the PR's tests kept, it aborts.
• Not tested: macOS, a console, any title.
• efd999c, a merge of main 938812d. The only conflict was docs/dev/debt/libkernel.md, where main reworded the sceKernelPread line next to this PR's close and getsockopt line; both changes are kept, and the PR's own diff is unchanged. On GitHub's runners (run), guest_sce_net, guest_socket_options, guest_socket_error and guest_socket_descriptor_flags pass 10 runs in a row on Linux and on Windows, 4 of 4, none skipped. The libc and libkernel suites pass on both (119 and 201 tests on Linux, 146 and 211 on Windows), run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
• abbb7ac, a merge of main 67bbe03. The only conflict was docs/dev/debt/libkernel.md, where feat(libkernel): answer sceKernelAddReadEvent and sceKernelAddWriteEvent on regular files #9099 (aarushkha, regular-file events) reworded the sceKernelAddReadEvent pipe entry and added its own regular-file entry just above this PR's close, setsockopt and getsockopt line; the merge keeps both of main's entries and this PR's line, and the PR's own diff is unchanged. On GitHub's runners (run), guest_sce_net, guest_socket_options, guest_socket_error and guest_socket_descriptor_flags pass 10 runs in a row with the change, on Linux and on Windows: 4 run, 4 pass, 0 skip, 0 fail. With main's code, guest_sce_net fails on both. The libc and libkernel suites pass on both (143 and 205 tests on Linux, 170 and 218 on Windows), run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
• fc47c2f, a merge of main a512780. The one conflict was in core/libs/tests/GuestSceNet.cpp, where refactor(libs): add full export headers for libs #10706 (the public-header migration) removed the test's local declarations, among them this PR's getsockopt_nid_postfix. The merge takes main's side, since libkernel's include/Api.hpp now declares getsockopt_nid_postfix with the same arguments. It also drops this PR's local sceNetGetsockopt declaration in libSceNet's Export.cpp, which tests/test_public_headers.py --audit-only reports as a second declaration next to Net.hpp's. So the PR's own diff loses those two declarations and is otherwise unchanged. On GitHub's runners (run), guest_sce_net, guest_socket_options, guest_socket_error, guest_socket_descriptor_flags and public_headers pass 10 runs in a row with the change, on Linux and on Windows: 5 run, 5 pass, 0 skip, 0 fail. With main's code, guest_sce_net fails on both. The libc and libkernel suites pass on both (143 and 206 tests on Linux, 170 and 222 on Windows), run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
• 00924d8, a merge of main 66e6772 (run). The one conflict was docs/dev/debt/libkernel.md's entry on libSceNet sockets, where main adds poll and fcntl and this PR adds getsockopt; the merge keeps both, in the same words as feat(libSceNet): implement the Ex and Internal socket forms, sceNetSysctl, scope-ID text and the default route #9578's merge. The PR's code diff is unchanged. guest_sce_net, guest_socket_descriptor_flags, guest_socket_error, guest_socket_options and public_headers pass 10 runs in a row. Linux: 5 run, 5 pass, 0 skip, 0 fail. Windows: 5 run, 5 pass, 0 skip, 0 fail. The before arm fails guest_sce_net on both. Suites: libc and libkernel pass on Linux (144 and 207) and on Windows (171 and 223). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: fix(libkernel): validate the pointers of getsockopt and setsockopt like FreeBSD #7511, docs(libkernel): correct three stale socket statements #9365, feat(libSceNet): read reuse and broadcast flags #9542 (after feat(libSceNet): read reuse and broadcast flags #9542, agreed in this thread; see Overlap)
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[MarcZX10]

The socket-hook change and the reported Windows/Linux tests look promising, but the checklist currently says Depends on: none while the PR description identifies two live overlaps. CONTRIBUTING.md requires overlap authors to agree on merge order and declare the dependency. The getsockopt_nid_postfix hunk overlaps open #7511, which validates pointer/length ordering in the same function; this PR inserts the libSceNet dispatch before that function’s current lookup. The docs/dev/debt/libkernel.md hunk overlaps open #9365 on the same socket-call sentence. Please coordinate with both authors, identify the agreed order, and update the checklist (and #9365’s “no open pull request edits these sentences” statement) before merge. I am not asking for a code change based on the reviewed diff.

[EagleBearerYS]

Thanks, @MarcZX10. This PR goes after both: Depends on now lists #7511 and #9365, and I've left a note on each. When it rebases on #7511, the libSceNet getter call moves after #7511's pointer and length checks, so a libSceNet socket fails those the same way libkernel's own sockets do. On #9365, this PR's addition of getsockopt to the close and setsockopt sentence goes on top of its correction, and the note there says that this PR also edits that sentence. The Overlap line in the description now gives the order.

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 4630	docs/dev/debt/libkernel.md
PR 5021	core/libs/prx/libkernel/Socket/src/Udp.cpp docs/dev/debt/libkernel.md
PR 5836	core/libs/prx/libSceNet/Export.cpp core/libs/tests/GuestSceNet.cpp
PR 6149	docs/dev/debt/libkernel.md
PR 6936	docs/dev/debt/libkernel.md
PR 7088	core/libs/prx/libkernel/Socket/src/Udp.cpp docs/dev/debt/libkernel.md
PR 7105	docs/dev/debt/libkernel.md
PR 7121	docs/dev/debt/libkernel.md
PR 7149	docs/dev/debt/libkernel.md
PR 7192	docs/dev/debt/libkernel.md
PR 7201	core/libs/prx/libSceNet/Export.cpp core/libs/prx/libkernel/Socket/src/Udp.cpp
PR 7447	docs/dev/debt/libkernel.md
PR 7509	docs/dev/debt/libkernel.md
PR 7804	docs/dev/debt/libkernel.md
PR 8605	core/libs/prx/libkernel/Socket/src/Udp.cpp docs/dev/debt/libkernel.md
PR 8725	docs/dev/debt/libkernel.md
PR 8916	docs/dev/debt/libkernel.md
PR 9156	docs/dev/debt/libkernel.md
PR 9199	docs/dev/debt/libkernel.md
PR 9284	docs/dev/debt/libkernel.md
PR 9293	docs/dev/debt/libkernel.md
PR 9297	docs/dev/debt/libkernel.md
PR 9302	docs/dev/debt/libkernel.md
PR 9326	docs/dev/debt/libkernel.md
PR 9389	docs/dev/debt/libkernel.md
PR 9442	docs/dev/debt/libkernel.md
PR 9557	docs/dev/debt/libkernel.md
PR 9566	docs/dev/debt/libkernel.md
PR 9575	docs/dev/debt/libkernel.md
PR 9713	core/libs/prx/libkernel/Socket/src/Udp.cpp
PR 9932	core/libs/prx/libSceNet/Export.cpp core/libs/tests/GuestSceNet.cpp
PR 9947	docs/dev/debt/libkernel.md
PR 9948	core/libs/prx/libSceNet/Export.cpp core/libs/prx/libkernel/Socket/include/SocketPoll.hpp core/libs/prx/libkernel/Socket/src/Udp.cpp core/libs/tests/GuestSceNet.cpp
PR 10319	docs/dev/debt/libkernel.md
PR 10442	docs/dev/debt/libkernel.md
PR 10552	docs/dev/debt/libkernel.md
PR 10558	docs/dev/debt/libkernel.md
PR 10749	docs/dev/debt/libkernel.md
PR 11002	docs/dev/debt/libkernel.md
PR 11085	docs/dev/debt/libkernel.md
PR 11215	docs/dev/debt/libkernel.md
PR 11223	core/libs/prx/libkernel/Socket/src/Udp.cpp docs/dev/debt/libkernel.md
PR 11264	docs/dev/debt/libkernel.md
PR 11267	docs/dev/debt/libkernel.md
PR 11272	core/libs/prx/libSceNet/Export.cpp core/libs/tests/GuestSceNet.cpp
PR 11341	core/libs/prx/libSceNet/Export.cpp docs/dev/debt/libkernel.md
PR 11403	docs/dev/debt/libkernel.md

[MarcZX10]

Re-reviewed the updated full body and diff after the author recorded the shared getsockopt/debt ordering with #7511 and #9365. The new getter registration is adjacent to the existing setter bridge, and the reported tests exercise the net-specific and libkernel dispatch paths with the FreeBSD timeval representation. Current hosted checks and the reported cross-platform test suites support the change; I did not run a local build. The prior overlap coordination blocker is resolved.

[MarcZX10]

Rechecked the current head and the updated dependency coordination. The PR now declares #7511 and #9365, explains how the getter dispatch follows #7511 pointer/length checks, and records the shared debt sentence order with #9365. The tests cover libSceNet boolean reads and libkernel timeval dispatch; author reports repeated Linux/Windows red/green runs and broader suites, with one Linux timing test failing on both base and change. Hosted eligibility, conventions, intake, labels and report checks pass; product/relinker jobs are skipped. I did not run local tests.

[xXPBSXx]

Linux data point for aea804d9fd: WSL2 Ubuntu 24.04 (virtualised kernel, not bare-metal), GCC 13.3.0, Release, BUILD_TESTING=ON, PR head alone (not merged onto main; its merge-base is a6dcea927).

• With the PR: ctest -R '^guest_sce_net$' passes 3 of 3.
• With libSceNet/Export.cpp, libkernel/Socket/include/SocketPoll.hpp, libkernel/Socket/src/Socket.cpp and libkernel/Socket/src/Udp.cpp restored to the merge-base (the PR's tests kept): fails (Subprocess aborted).
  Not tested: macOS, Windows, a title run.
  AI-assisted (Claude Code).

[EagleBearerYS]

@xXPBSXx thanks for the Linux run, both red and green. It's in Tested now.

[Queuse]

@EagleBearerYS I found a partial overlap with my #9542 while checking the wider socket-option queue: both add native SO_REUSEADDR/SO_BROADCAST readback. Mine also handles short output buffers and has a dedicated regression. #9578 carries your readback commit as well.

Would #9523 → #9542 → this PR → #9578 work for you, keeping your existing #7511/#9365 prerequisites? That would let this PR keep the getter forwarding and its tests while dropping the duplicate readback hunk. If yours needs to land first, I can instead rebase #9542 and retain only its short-buffer behavior and dedicated test. I have marked mine as a draft while we agree on the order.

[EagleBearerYS]

@Queuse that order works: #9523 → #9542 → this PR → #9578, with #7511 and #9365 still before this PR. #9542's readback is the same as this PR's e887616 (the host's SO_REUSEADDR and SO_BROADCAST, returned as 0x4 and 0x20), and it adds the short-buffer handling. The two don't conflict as text, but together they'd put the same two case labels in sceNetGetsockopt's switch, so one has to go. Once #9542 lands, this PR drops that hunk when it next merges main and keeps its getter forwarding and its tests. #9578, which carries the same commit, follows at its own next merge. I checked the composition, #9542 with this PR merged in and that hunk dropped: guest_sce_net and guest_net_boolean_readback pass on both runners (run). With that, the order is agreed, and I've added it to both PRs' Depends on. #9542 can go back to ready whenever you like.

[Queuse]

Thanks, @EagleBearerYS. I've recorded the agreed #9523 → #9542 → #9515 → #9578 order and your combined test results in #9542, keeping your #7511/#9365 prerequisites explicit. The overlap item in the checklist is now settled. #9542 is still a draft while the separate test-reorganization coordination with #6044 is pending.

[oneandonlydean]

@EagleBearerYS This PR now conflicts with main in core/libs/tests/GuestSceNet.cpp. Main’s #10706 public-header migration removed local socket declarations, while this PR’s test still supplies GuestPollEntry and a local poll declaration. Main now requires each exported function to be declared once, in the library's public header (checked by the public_headers tests). Could you reconcile the poll test’s argument types with the current public interface?

[EagleBearerYS]

@oneandonlydean Thanks, merged main (a512780) in fc47c2f. The conflict was the test's local declaration block, which #10706 removed. This PR had added only getsockopt_nid_postfix there, and libkernel's include/Api.hpp now declares it with the same arguments, so the merge takes main's side. GuestPollEntry and the poll declaration in that block were main's lines, not this PR's, and they go with it. The merge also drops this PR's local sceNetGetsockopt declaration in libSceNet's Export.cpp, which public_headers reported as a second declaration next to Net.hpp's. guest_sce_net, the three guest_socket_* tests and public_headers pass 10 runs in a row on Linux and Windows (run). #9578, which builds on this PR, has the same merge in 6c938db.

[Queuse]

The #9523/#9542 branches are refreshed, and #9542 is ready for review. The agreed #9523 → #9542 → #9515 → #9578 order still works on my side. #6044 was closed without the test relocation, so the draft blocker mentioned in my earlier reply is gone.

I preserved main's merged protocol/TCP_NODELAY getters and their short-output rules in both refreshes; each branch passes its own regression plus the protocol, TCP_NODELAY and buffer controls on Windows.