[Retr0G28]

What

fstat on a socketpair end reported st_dev and st_ino 0, which the debt entry recorded. FreeBSD 11's uipc_sense, the pru_sense of a local socket, sets st_dev to NODEV and st_ino to the socket's unp_ino. A socket gets its unp_ino from one global counter the first time it is sensed, skipping 0, and keeps it. dev_t and ino_t are 32-bit in FreeBSD 11, as in FileStat, so NODEV is 0xFFFFFFFF. Inet sockets use pru_sense_null, which leaves both 0.

GuestSockets::Stat now does the same for a socket of family AF_UNIX, which only socketpair creates. The socket gets a number from a counter under socketsMutex the first time it is stat'ed, and the number is kept in Socket. A dup shares the Socket, so it reports the same number, as it shares the struct socket on FreeBSD. The other fields are unchanged.

main added a sentence to the fstat on a guest socket debt entry saying a socketpair end reports st_dev and st_ino 0. It now names only what still differs: the number comes from a counter this process starts at 1, where FreeBSD's unp_ino is one counter for every local socket on the system, so the numbers differ from the console's. st_dev now matches.

Test: guest_socket_pair checks that both ends of a stream pair report st_dev 0xFFFFFFFF and a nonzero st_ino. The two ends must differ, a second fstat or sceKernelFstat must give the same number, and a dup of an end must report that end's number. An AF_INET socket still reports st_dev and st_ino 0.

Tested

Windows 11, WinLibs MinGW-w64 GCC 15.2.0 r7, CMake + Ninja, Release, -DBUILD_TESTING=ON, rebased on main 6cd36e6.

• Before the rebase, on main a23f8fc: with main's Udp.cpp, guest_socket_pair fails (exit 0xc0000409 from the failed Require). With this PR it passes, and ctest -R "socket|sendfile|kqueue|poll" passes 39 of 39.
• After the rebase: ctest -R "socket|recvmsg|sendfile|kqueue|^guest_poll$" passes 32 of 32 when run serially.
• After merging main afc780e (head ada29c7): the same pattern passes 42 of 42 when run serially.
• python tools/check_conventions.py --base origin/main reports no errors.
• Not run on Linux or macOS, where the pair is a host AF_UNIX pair and the same code path runs. No title or homebrew run.

Overlap (tools/pr_overlap.py):

• fix(libkernel): return ENOTCONN from getpeername once a socketpair peer closed #8344 (mine) added the peer member next to the new inode member (Udp.cpp L124). It merged first; this branch keeps both members.
• fix(libkernel): clear fstat read bits on a socket after the peer closes #8605 changes the same fstat on a guest socket debt entry, for the read bits after the peer closes, and code near GuestSockets::Stat. The two touch different sentences and fields, so the second needs only a rebase of that entry.
• fix(libkernel): poll kqueues as kqueue_poll answers #7605 and docs(libkernel): correct three stale socket statements #9365 change the debt lines next to it, the poll entry and stale socket statements. Neither changes the st_dev or st_ino sentence.
• None of the PRs it could not compare changes GuestSockets::Stat or the socketpair fstat checks.

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: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[MarcZX10]

The GuestSockets::Stat change and tests look consistent with the described AF_UNIX behavior: inode assignment occurs under socketsMutex, the Socket owns the value so dup shares it, and the AF_INET control preserves zeros. One coordination item remains in the PR body: it says #8344 adds a Socket member adjacent to this new inode field and whichever merges second needs a rebase, but the checklist says Depends on: none. CONTRIBUTING.md asks overlapping PRs to agree on order and declare that dependency. Please record the agreed order (or state that #8344 is the prerequisite) so reviewers and automation can track the shared insertion hunk. I have not independently run the Windows socket tests; hosted product build legs were skipped by eligibility.

[Retr0G28]

@MarcZX10 Thanks. Both PRs are mine. The order is #8344 first, so the checklist now says Depends on: #8344 and the overlap note says so too. After #8344 merges I will merge main into this branch, keep both Socket members, and rerun guest_socket_pair and the socket tests.

[MarcZX10]

The author has resolved the coordination item: #8344 is explicitly first, and this PR now declares Depends on: #8344 and says it will rebase and rerun guest_socket_pair after #8344 merges. I approve the reviewed implementation and tests at head c27cdcaf4cc1218a0679720ad6f3200645599255. The previously reviewed behavior remains consistent: inode assignment is protected by socketsMutex, lifetime is owned by the shared Socket (so dup retains identity), and AF_INET stats remain zero. Hosted eligibility/conventions/intake/labels pass; product checks were skipped. Windows socket test results are author-reported, not independently reproduced. The PR remains dependent on #8344 and should be revalidated after that rebase.

[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 core/libs/tests/GuestSocketPair.cpp docs/dev/debt/libkernel.md
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/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 core/libs/tests/GuestSocketPair.cpp 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 9389	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 9947	docs/dev/debt/libkernel.md
PR 9948	core/libs/prx/libkernel/Socket/src/Udp.cpp
PR 10442	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 core/libs/tests/GuestSocketPair.cpp docs/dev/debt/libkernel.md
PR 11264	docs/dev/debt/libkernel.md
PR 11267	docs/dev/debt/libkernel.md
PR 11341	docs/dev/debt/libkernel.md

[Retr0G28]

Merged main f26434a to clear the same GuestSocketPair.cpp declaration conflict as #8344. The diff against main is unchanged. Head e7de2a5, Windows 11, GCC 15.2.0: 41 of 41 socket, recvmsg, sendfile, kqueue and poll tests pass when run serially.

[oneandonlydean]

@Retr0G28 thanks for the socket inode work. #10621 conflicts with main in two places.

The Udp.cpp hunk is trivial: both std::uint32_t inode and main's new std::weak_ptr<Socket> peer stay in the struct. The debt entry for fstat on a guest socket in docs/dev/debt/libkernel.md needs your input, though.

Main added this sentence at the end of that entry: "A socketpair end reports st_dev and st_ino 0, where FreeBSD's uipc_sense reports NODEV and a per-socket number". Your PR now gives unix-family sockets a lazily assigned st_ino, so the st_ino 0 half is no longer true for them after your change.

Could you merge main and restate that sentence to what still differs from FreeBSD once your PR is in? For example, whether st_dev is still 0 and how the number differs from uipc_sense's. Please keep main's wording elsewhere in the entry, including sceKernelClose in the first bullet.

[Retr0G28]

@oneandonlydean Merged main afc780e. In Udp.cpp both std::uint32_t inode and std::weak_ptr<Socket> peer stay in the struct. The fstat on a guest socket entry keeps main's wording, sceKernelClose included; only the last sentence changes.

After this PR, st_dev is NODEV (0xFFFFFFFF) as in uipc_sense, so it's no longer a gap. st_ino is a nonzero per-socket number assigned on the first fstat and kept, also as in uipc_sense. What still differs is where the number comes from, so the sentence now reads:

A socketpair end's st_ino comes from a counter this process starts at 1, where FreeBSD's uipc_sense uses one counter for every local socket on the system, so the numbers differ from the console's

Head ada29c7, Windows 11 (GCC 15.2.0): ctest -R "socket|recvmsg|sendfile|kqueue|^guest_poll$" passes 42 of 42 when run serially, and check_conventions.py passes. Depends on is now none, since #8344 has merged.